{"task_id": "knows.docs_11_personal_recipe_ocr.1", "short_id": "11a", "family": "docs_11_personal_recipe_ocr", "instance": "instance_1", "instance_num": 1, "artifact": "Docs", "group": "Personal", "domain": "Recipes", "prompt": "Create a recipe from the Recipe Coral template in Google Docs. First, replace the name of the recipe with the correct recipe name by looking at the image saved in the drive at https://drive.google.com/file/d/1S7XtioiuOZXdNt6fSS_9FkGQ42UDZ8l-/view?usp=sharing. Then, copy the ingredient list into the template and paste the instructions into the corresponding field. Get tips from an online source, and make sure to cite source urls for tips below the tips entries. Use direct quotes for all tips. Replace the default image with a cropped image from the original drive link that shows the recipe text clearly. Fill out the rest of the template with relevant information from the recipe. Then search the internet for at least 3 similar recipes and put them on separate pages with the same format as the first recipe. Put the link for the source of each additional recipe on a line below the title fitting with the formatting. Make sure each page of recipes is visually distinct from the others. Finally, check the other online recipes and fill in the tips with some useful information that you found.", "checkpoints": [{"n": 1, "title": "Original Recipe Page", "description": "", "steps": [{"name": null, "description": "Title of the first page has been changed to \"Pumpkin Soup\"", "text": "Title of the first page has been changed to \"Pumpkin Soup\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ingredients of the first page match the picture ingredient list", "text": "Ingredients of the first page match the picture ingredient list", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Preparation steps match picture preparation steps", "text": "Preparation steps match picture preparation steps", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Tips section has relevant tips to making the recipe", "text": "Tips section has relevant tips to making the recipe", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has valid source urls linked below the tips text", "text": "Tips section has valid source urls linked below the tips text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tips are all exclusively direct quotes from source url", "text": "Tips are all exclusively direct quotes from source url", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Image is of the recipe from the original image", "text": "Image is of the recipe from the original image", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is cropped to properly fit the template", "text": "Image is cropped to properly fit the template", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image shows the recipe text clearly", "text": "Image shows the recipe text clearly", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "text": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 2, "title": "Additional Recipe Pages Formatting (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a recipe title related to fall, soups, pumpkins, or thanksgiving.", "text": "Each additional recipe has a recipe title related to fall, soups, pumpkins, or thanksgiving.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has a valid source url for the recipe below the title", "text": "Each additional recipe has a valid source url for the recipe below the title", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "text": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe ingredient content is modified from the default", "text": "Each additional recipe ingredient content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe preparation content is modified from the default", "text": "Each additional recipe preparation content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe tips content is modified from the default", "text": "Each additional recipe tips content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "For each additional recipe information under photo is modified from the default", "text": "For each additional recipe information under photo is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe is visually distinguishable, via color or font changes to text.", "text": "Each additional recipe is visually distinguishable, via color or font changes to text.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Additional Recipe Pages Content (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a photo from the source url of the item in the recipe", "text": "Each additional recipe has a photo from the source url of the item in the recipe", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has ingredients list which substantially matches the source ingredients", "text": "Each additional recipe has ingredients list which substantially matches the source ingredients", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has preparation steps which substantially match the source steps", "text": "Each additional recipe has preparation steps which substantially match the source steps", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has tips which are clearly extracted from the source", "text": "Each additional recipe has tips which are clearly extracted from the source", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "text": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Websites Visited Check", "description": "All relevant source URLs were visited", "steps": [{"name": null, "description": "First page tips source urls are in the browsing history of the agent", "text": "First page tips source urls are in the browsing history of the agent", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Additional recipe source urls are in the browsing history of the agent (x3)", "text": "Additional recipe source urls are in the browsing history of the agent (x3)", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 25, "total_points": 53, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/file/d/1S7XtioiuOZXdNt6fSS_9FkGQ42UDZ8l-/view?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_1/checkpoints.md"}
{"task_id": "knows.docs_11_personal_recipe_ocr.2", "short_id": "11b", "family": "docs_11_personal_recipe_ocr", "instance": "instance_2", "instance_num": 2, "artifact": "Docs", "group": "Personal", "domain": "Recipes", "prompt": "Create a recipe from the Recipe Coral template in Google Docs. First, replace the name of the recipe with the correct recipe name by looking at the image saved in the drive at https://drive.google.com/drive/folders/1dNvmuIo5L2952H3NSEuHtYDkQZau1hmq?usp=sharing. Then, copy the ingredient list into the template and extract any preparation instructions into the corresponding field. Get tips from an online source, and make sure to cite source urls for tips below the tips entries. Use direct quotes for all tips. Replace the default image with the recipe image from the original drive link that shows the recipe text clearly. Fill out the rest of the template with relevant information from the recipe. Then search the internet for at least 3 similar recipes and put them on separate pages with the same format as the first recipe. Put the link for the source of each additional recipe on a line below the title fitting with the formatting. Make sure each page of recipes is visually distinct from the others. Finally, check the other online recipes and fill in the tips with some useful information that you found.", "checkpoints": [{"n": 1, "title": "Original Recipe Page", "description": "", "steps": [{"name": null, "description": "Title of the first page has been changed to \"Healthy Pizza\"", "text": "Title of the first page has been changed to \"Healthy Pizza\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ingredients of the first page match the picture ingredient list", "text": "Ingredients of the first page match the picture ingredient list", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Preparation steps match picture preparation steps", "text": "Preparation steps match picture preparation steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has relevant tips to making the recipe", "text": "Tips section has relevant tips to making the recipe", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has valid source urls linked below the tips text", "text": "Tips section has valid source urls linked below the tips text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tips are all exclusively direct quotes from source url", "text": "Tips are all exclusively direct quotes from source url", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Image is of the recipe from the original image", "text": "Image is of the recipe from the original image", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is sized to properly fit the template", "text": "Image is sized to properly fit the template", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image shows the recipe text clearly", "text": "Image shows the recipe text clearly", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "text": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 2, "title": "Additional Recipe Pages Formatting (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a recipe title related to pizza.", "text": "Each additional recipe has a recipe title related to pizza.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has a valid source url for the recipe below the title", "text": "Each additional recipe has a valid source url for the recipe below the title", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "text": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe ingredient content is modified from the default", "text": "Each additional recipe ingredient content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe preparation content is modified from the default", "text": "Each additional recipe preparation content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe tips content is modified from the default", "text": "Each additional recipe tips content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "For each additional recipe information under photo is modified from the default", "text": "For each additional recipe information under photo is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe is visually distinguishable, via color or font changes to text.", "text": "Each additional recipe is visually distinguishable, via color or font changes to text.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Additional Recipe Pages Content (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a photo from the source url of the item in the recipe", "text": "Each additional recipe has a photo from the source url of the item in the recipe", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has ingredients list which substantially matches the source ingredients", "text": "Each additional recipe has ingredients list which substantially matches the source ingredients", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has preparation steps which substantially match the source steps", "text": "Each additional recipe has preparation steps which substantially match the source steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has tips which are clearly extracted from the source", "text": "Each additional recipe has tips which are clearly extracted from the source", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "text": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Websites Visited Check", "description": "All relevant source URLs were visited", "steps": [{"name": null, "description": "First page tips source urls are in the browsing history of the agent", "text": "First page tips source urls are in the browsing history of the agent", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Additional recipe source urls are in the browsing history of the agent (x3)", "text": "Additional recipe source urls are in the browsing history of the agent (x3)", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 25, "total_points": 53, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1dNvmuIo5L2952H3NSEuHtYDkQZau1hmq?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_2/checkpoints.md"}
{"task_id": "knows.docs_11_personal_recipe_ocr.3", "short_id": "11c", "family": "docs_11_personal_recipe_ocr", "instance": "instance_3", "instance_num": 3, "artifact": "Docs", "group": "Personal", "domain": "Recipes", "prompt": "Create a recipe from the Recipe Coral template in Google Docs. First, replace the name of the recipe with the correct recipe name by looking at the image saved in the drive at https://drive.google.com/drive/folders/1fjJwG19JgIE3_hAmcg0TMa_bkFvv9tlN?usp=sharing. Then, copy the ingredient list into the template and include any preparation instruction in the corresponding section. Get tips from an online source, and make sure to cite source urls for tips below the tips entries. Use direct quotes for all tips. Replace the default image with a cropped image from the original drive link that shows the recipe text clearly. Fill out the rest of the template with relevant information from the recipe. Then search the internet for at least 4 similar recipes and put them on separate pages with the same format as the first recipe. Put the link for the source of each additional recipe on a line below the title fitting with the formatting. Make sure each page of recipes is visually distinct in font or heading colors from the others. Finally, check the other online recipes and fill in the tips with some useful information that you found.", "checkpoints": [{"n": 1, "title": "Original Recipe Page", "description": "", "steps": [{"name": null, "description": "Title of the first page has been changed to \"Larry's Favorite Cookies\"", "text": "Title of the first page has been changed to \"Larry's Favorite Cookies\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ingredients of the first page match the picture ingredient list", "text": "Ingredients of the first page match the picture ingredient list", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Preparation steps are either blank or mention no preparation steps found", "text": "Preparation steps are either blank or mention no preparation steps found", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Tips section has relevant tips to making the recipe", "text": "Tips section has relevant tips to making the recipe", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has valid source urls linked below the tips text", "text": "Tips section has valid source urls linked below the tips text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tips are all exclusively direct quotes from source url", "text": "Tips are all exclusively direct quotes from source url", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Image is of the recipe from the original image", "text": "Image is of the recipe from the original image", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is cropped to properly fit the template", "text": "Image is cropped to properly fit the template", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image shows the recipe text clearly", "text": "Image shows the recipe text clearly", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "text": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 2, "title": "Additional Recipe Pages Formatting (x4)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a recipe title related to cookies or someones favorite dessert.", "text": "Each additional recipe has a recipe title related to cookies or someones favorite dessert.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has a valid source url for the recipe below the title", "text": "Each additional recipe has a valid source url for the recipe below the title", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "text": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe ingredient content is modified from the default", "text": "Each additional recipe ingredient content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe preparation content is modified from the default", "text": "Each additional recipe preparation content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe tips content is modified from the default", "text": "Each additional recipe tips content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "For each additional recipe information under photo is modified from the default", "text": "For each additional recipe information under photo is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe is visually distinguishable, via color or font changes to text.", "text": "Each additional recipe is visually distinguishable, via color or font changes to text.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Additional Recipe Pages Content (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a photo from the source url of the item in the recipe", "text": "Each additional recipe has a photo from the source url of the item in the recipe", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has ingredients list which substantially matches the source ingredients", "text": "Each additional recipe has ingredients list which substantially matches the source ingredients", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has preparation steps which substantially match the source steps", "text": "Each additional recipe has preparation steps which substantially match the source steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has tips which are clearly extracted from the source", "text": "Each additional recipe has tips which are clearly extracted from the source", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "text": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Websites Visited Check", "description": "All relevant source URLs were visited", "steps": [{"name": null, "description": "First page tips source urls are in the browsing history of the agent", "text": "First page tips source urls are in the browsing history of the agent", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Additional recipe source urls are in the browsing history of the agent (x3)", "text": "Additional recipe source urls are in the browsing history of the agent (x3)", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 25, "total_points": 61, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1fjJwG19JgIE3_hAmcg0TMa_bkFvv9tlN?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_3/checkpoints.md"}
{"task_id": "knows.docs_11_personal_recipe_ocr.4", "short_id": "11d", "family": "docs_11_personal_recipe_ocr", "instance": "instance_4", "instance_num": 4, "artifact": "Docs", "group": "Personal", "domain": "Recipes", "prompt": "Create a recipe from the Recipe Coral template in Google Docs. First, replace the name of the recipe with the correct recipe name by looking at the image saved in the drive at https://drive.google.com/drive/folders/11wZ0Bbm_FpNmmVaBe5doN_0ICd30GxBN?usp=sharing. Then, copy the ingredient list into the template and include any preparation instruction in the corresponding section. Get tips from an online source, and make sure to cite source urls for tips below the tips entries. Use direct quotes for all tips. Replace the default image with a cropped image from the original drive link that shows the recipe text clearly. Fill out the rest of the template with relevant information from the recipe. Then search the internet for exactly 3 similar recipes and put them on separate pages with the same format as the first recipe. Put the link for the source of each additional recipe on a line below the title fitting with the formatting. Make sure each page of recipes is visually distinct in font or heading colors from the others. Finally, check the other online recipes and fill in the tips with some useful information that you found.", "checkpoints": [{"n": 1, "title": "Original Recipe Page", "description": "", "steps": [{"name": null, "description": "Title of the first page has been changed to \"Cardamom Cookies\"", "text": "Title of the first page has been changed to \"Cardamom Cookies\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ingredients of the first page match the picture ingredient list", "text": "Ingredients of the first page match the picture ingredient list", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Preparation steps are either blank, mention no preparation steps found, or list the one step from the ingredients", "text": "Preparation steps are either blank, mention no preparation steps found, or list the one step from the ingredients", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Tips section has relevant tips to making the recipe", "text": "Tips section has relevant tips to making the recipe", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has valid source urls linked below the tips text", "text": "Tips section has valid source urls linked below the tips text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tips are all exclusively direct quotes from source url", "text": "Tips are all exclusively direct quotes from source url", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Image is of the recipe from the original image", "text": "Image is of the recipe from the original image", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is cropped to properly fit the template", "text": "Image is cropped to properly fit the template", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image shows the recipe text clearly", "text": "Image shows the recipe text clearly", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "text": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 2, "title": "Additional Recipe Pages Formatting (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a recipe title related to cookies.", "text": "Each additional recipe has a recipe title related to cookies.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has a valid source url for the recipe below the title", "text": "Each additional recipe has a valid source url for the recipe below the title", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "text": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe ingredient content is modified from the default", "text": "Each additional recipe ingredient content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe preparation content is modified from the default", "text": "Each additional recipe preparation content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe tips content is modified from the default", "text": "Each additional recipe tips content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "For each additional recipe information under photo is modified from the default", "text": "For each additional recipe information under photo is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe is visually distinguishable, via color or font changes to text.", "text": "Each additional recipe is visually distinguishable, via color or font changes to text.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Additional Recipe Pages Content (x3)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a photo from the source url of the item in the recipe", "text": "Each additional recipe has a photo from the source url of the item in the recipe", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has ingredients list which substantially matches the source ingredients", "text": "Each additional recipe has ingredients list which substantially matches the source ingredients", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has preparation steps which substantially match the source steps", "text": "Each additional recipe has preparation steps which substantially match the source steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has tips which are clearly extracted from the source", "text": "Each additional recipe has tips which are clearly extracted from the source", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "text": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Websites Visited Check", "description": "All relevant source URLs were visited", "steps": [{"name": null, "description": "First page tips source urls are in the browsing history of the agent", "text": "First page tips source urls are in the browsing history of the agent", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Additional recipe source urls are in the browsing history of the agent (x3)", "text": "Additional recipe source urls are in the browsing history of the agent (x3)", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 25, "total_points": 53, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/11wZ0Bbm_FpNmmVaBe5doN_0ICd30GxBN?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_4/checkpoints.md"}
{"task_id": "knows.docs_11_personal_recipe_ocr.5", "short_id": "11e", "family": "docs_11_personal_recipe_ocr", "instance": "instance_5", "instance_num": 5, "artifact": "Docs", "group": "Personal", "domain": "Recipes", "prompt": "Create a recipe from the Recipe Coral template in Google Docs. First, replace the name of the recipe with the correct recipe name by looking at the images saved in the drive at https://drive.google.com/drive/folders/1Ekh0zoN6tNq_rBGsacGqrxunUeOpdKsU?usp=sharing. Then, copy the ingredient list into the template and include any preparation instructions in the corresponding section. Get tips from the recipe and an online source, and make sure to cite source urls for tips below the tips entries. Use direct quotes for all tips. Replace the default image with a cropped image from the original drive link that shows the recipe ingredient text clearly. Fill out the rest of the template with relevant information from the recipe. Then search the internet for at least 5 similar recipes and put them on separate pages with the same format as the first recipe. Put the link for the source of each additional recipe on a line below the title fitting with the formatting. Make sure each page of recipes is visually distinct in font or heading colors from the others. Finally, check the other online recipes and fill in the tips with some useful information that you found.", "checkpoints": [{"n": 1, "title": "Original Recipe Page", "description": "", "steps": [{"name": null, "description": "Title of the first page has been changed to \"Pasta Puttanesca for Dez\"", "text": "Title of the first page has been changed to \"Pasta Puttanesca for Dez\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ingredients of the first page match the picture ingredient list", "text": "Ingredients of the first page match the picture ingredient list", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Preparation steps match picture preparation steps", "text": "Preparation steps match picture preparation steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has relevant tips to making the recipe, including picture tips", "text": "Tips section has relevant tips to making the recipe, including picture tips", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Tips section has valid source urls linked below the tips text", "text": "Tips section has valid source urls linked below the tips text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tips are all exclusively direct quotes from source url or the picture", "text": "Tips are all exclusively direct quotes from source url or the picture", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Image is of the recipe from the original image", "text": "Image is of the recipe from the original image", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is cropped to properly fit the template", "text": "Image is cropped to properly fit the template", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image shows the recipe text clearly", "text": "Image shows the recipe text clearly", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "text": "Information under photo (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 2, "title": "Additional Recipe Pages Formatting (x5)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a recipe title related to Pasta or Puttanesca.", "text": "Each additional recipe has a recipe title related to Pasta or Puttanesca.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has a valid source url for the recipe below the title", "text": "Each additional recipe has a valid source url for the recipe below the title", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "text": "Each additional recipe has a format matching the first page (2 column, title, picture, ready time, serves and calories on the left, Ingredients, preparation, and tips on the right.)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe ingredient content is modified from the default", "text": "Each additional recipe ingredient content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe preparation content is modified from the default", "text": "Each additional recipe preparation content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe tips content is modified from the default", "text": "Each additional recipe tips content is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "For each additional recipe information under photo is modified from the default", "text": "For each additional recipe information under photo is modified from the default", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe is visually distinguishable, via color or font changes to text.", "text": "Each additional recipe is visually distinguishable, via color or font changes to text.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Additional Recipe Pages Content (x5)", "description": "", "steps": [{"name": null, "description": "Each additional recipe has a photo from the source url of the item in the recipe", "text": "Each additional recipe has a photo from the source url of the item in the recipe", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has ingredients list which substantially matches the source ingredients", "text": "Each additional recipe has ingredients list which substantially matches the source ingredients", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has preparation steps which substantially match the source steps", "text": "Each additional recipe has preparation steps which substantially match the source steps", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each additional recipe has tips which are clearly extracted from the source", "text": "Each additional recipe has tips which are clearly extracted from the source", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "text": "Each additional recipe has information under the photo relevant to the source recipe (Ready in, serves and calories is modified based on recipe info)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Websites Visited Check", "description": "All relevant source URLs were visited", "steps": [{"name": null, "description": "First page tips source urls are in the browsing history of the agent", "text": "First page tips source urls are in the browsing history of the agent", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Additional recipe source urls are in the browsing history of the agent (x5)", "text": "Additional recipe source urls are in the browsing history of the agent (x5)", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 25, "total_points": 81, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1Ekh0zoN6tNq_rBGsacGqrxunUeOpdKsU?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_11_personal_recipe_ocr/instance_5/checkpoints.md"}
{"task_id": "knows.docs_1_formal_letter.1", "short_id": "1b", "family": "docs_1_formal_letter", "instance": "instance_1", "instance_num": 1, "artifact": "Docs", "group": "Letters", "domain": "Business Letter", "prompt": "Structure a formal letter written in my name using Google Docs. \nUse my website for my info: https://ethanashby.github.io/.\nInclude my employers logo in the upper left corner if available online.\nInclude the following information in the letter header:\n- My Name\n- My Email Address\n- My Job/Position Title\n\nTake the photo of my signature from the drive folder here: https://drive.google.com/drive/folders/1kJM2XLnf8qIu652O2H_uk23i6Es9Ajgf?usp=drive_link and add it to the end of the letter.", "checkpoints": [{"n": 1, "title": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "description": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "steps": [{"name": null, "description": "Exact match of the name to the ground-truth name.", "text": "Exact match of the name to the ground-truth name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the name to the upper-left or upper-right of the document.", "text": "Location match of the name to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the email to the ground-truth email.", "text": "Exact match of the email to the ground-truth email.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the email to the upper-left or upper-right of the document.", "text": "Location match of the email to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the job/position title to the ground-truth job/position title.", "text": "Exact match of the job/position title to the ground-truth job/position title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the job/position title to the upper-left or upper-right of the document.", "text": "Location match of the job/position title to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The logo was added to the requested place in the document.", "description": "The logo was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an image of the University of Washington.", "text": "Image in the doc is an image of the University of Washington.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is placed at the top left of the document", "text": "Logo is placed at the top left of the document", "category": "spatial_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The signature was added to the requested place in the document.", "description": "The signature was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an exact match to the ground-truth signature.", "text": "Image in the doc is an exact match to the ground-truth signature.", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "text": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 10, "total_points": 10, "requires_browsing_history": false, "source_assets": ["https://drive.google.com/drive/folders/1kJM2XLnf8qIu652O2H_uk23i6Es9Ajgf?usp=drive_link"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_1/checkpoints.md"}
{"task_id": "knows.docs_1_formal_letter.2", "short_id": "1a", "family": "docs_1_formal_letter", "instance": "instance_2", "instance_num": 2, "artifact": "Docs", "group": "Letters", "domain": "Business Letter", "prompt": "Structure a formal letter written in my name using Google Docs. \nUse my website for my info: https://www.andrewreischling.com/.\nInclude my employers logo in the upper left corner if available online.\nInclude the following information in the letter header:\n- My Name\n- My Email Address\n- My Job/Position Title\n\nTake the photo of my signature from the drive folder here: https://drive.google.com/drive/folders/1-DEfT1Sf1Qhfh3LftXXaVjP49T-xhEU2?usp=sharing and add it to the end of the letter.\nSave the completed letter as \"<first_name>_<last_name>_formal_letter\" in the \"WebAgentsShared -> Task Outputs\" directory in Google Drive.", "checkpoints": [{"n": 1, "title": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "description": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "steps": [{"name": null, "description": "Exact match of the name to the ground-truth name.", "text": "Exact match of the name to the ground-truth name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the name to the upper-left or upper-right of the document.", "text": "Location match of the name to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the email to the ground-truth email.", "text": "Exact match of the email to the ground-truth email.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the email to the upper-left or upper-right of the document.", "text": "Location match of the email to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the job/position title to the ground-truth job/position title.", "text": "Exact match of the job/position title to the ground-truth job/position title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the job/position title to the upper-left or upper-right of the document.", "text": "Location match of the job/position title to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The logo was added to the requested place in the document.", "description": "The logo was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an image of the Ideo logo.", "text": "Image in the doc is an image of the Ideo logo.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is placed at the top left of the document", "text": "Logo is placed at the top left of the document", "category": "spatial_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The signature was added to the requested place in the document.", "description": "The signature was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an exact match to the ground-truth signature.", "text": "Image in the doc is an exact match to the ground-truth signature.", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "text": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 10, "total_points": 10, "requires_browsing_history": false, "source_assets": ["https://drive.google.com/drive/folders/1-DEfT1Sf1Qhfh3LftXXaVjP49T-xhEU2?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_2/checkpoints.md"}
{"task_id": "knows.docs_1_formal_letter.3", "short_id": "1c", "family": "docs_1_formal_letter", "instance": "instance_3", "instance_num": 3, "artifact": "Docs", "group": "Letters", "domain": "Business Letter", "prompt": "Structure a formal letter written in my name using Google Docs.\nUse my website for my info: https://elisakreiss.com/.\nInclude my employer's logo in the upper left corner if available online. Use the first affiliation listed if multiple exist. For incoming positions, use the future institution.\nInclude the following information in the letter header:\n\n- My Name\n- My Job/Position Title\n- My Institution/Company\n- My Email Address\n\nTake the photo of my signature from the drive folder here: https://drive.google.com/drive/folders/1LatG8gzruM50ebGfosWw49xXGxLQZRMW and add it to the end of the letter.\nSave the completed letter as \"<first_name>\\_<last_name>\\_formal_letter\" in the \"WebAgentsShared -> Task Outputs\" directory in Google Drive.", "checkpoints": [{"n": 1, "title": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "description": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "steps": [{"name": null, "description": "Exact match of the name to the ground-truth name.", "text": "Exact match of the name to the ground-truth name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the name to the upper-left or upper-right of the document.", "text": "Location match of the name to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the job/position title to the ground-truth job/position title.", "text": "Exact match of the job/position title to the ground-truth job/position title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the job/position title to the upper-left or upper-right of the document.", "text": "Location match of the job/position title to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the institution/company to the ground-truth institution/company.", "text": "Exact match of the institution/company to the ground-truth institution/company.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the institution/company to the upper-left or upper-right of the document", "text": "Location match of the institution/company to the upper-left or upper-right of the document", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the email to the ground-truth email.", "text": "Exact match of the email to the ground-truth email.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the email to the upper-left or upper-right of the document.", "text": "Location match of the email to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The logo was added to the requested place in the document.", "description": "The logo was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an image of UCLA or Coalas Lab.", "text": "Image in the doc is an image of UCLA or Coalas Lab.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is placed at the top left of the document", "text": "Logo is placed at the top left of the document", "category": "spatial_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The signature was added to the requested place in the document.", "description": "The signature was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an exact match to the ground-truth signature.", "text": "Image in the doc is an exact match to the ground-truth signature.", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "text": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 12, "total_points": 12, "requires_browsing_history": false, "source_assets": ["https://drive.google.com/drive/folders/1LatG8gzruM50ebGfosWw49xXGxLQZRMW"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_3/checkpoints.md"}
{"task_id": "knows.docs_1_formal_letter.4", "short_id": "1d", "family": "docs_1_formal_letter", "instance": "instance_4", "instance_num": 4, "artifact": "Docs", "group": "Letters", "domain": "Business Letter", "prompt": "Structure a formal letter written in my name using Google Docs.\nUse my website for my info: https://nickatomlin.github.io/.\nInclude my employer's logo in the upper left corner if available online. Use the first affiliation listed if multiple exist. For incoming positions, use the future institution.\nInclude the following information in the letter header:\n\n- My Name\n- My Email Address\n- My Job/Position Title\n- My Institution/Company\n\nTake the photo of my signature from the drive folder here: https://drive.google.com/drive/folders/1ykj346THKjAlKUDUNHHZD0_C84PUASp7 and add it to the end of the letter.\nSave the completed letter as \"<first_name>\\_<last_name>\\_formal_letter\" in the \"WebAgentsShared -> Task Outputs\" directory in Google Drive.", "checkpoints": [{"n": 1, "title": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "description": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "steps": [{"name": null, "description": "Exact match of the name to the ground-truth name.", "text": "Exact match of the name to the ground-truth name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the name to the upper-left or upper-right of the document.", "text": "Location match of the name to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the email to the ground-truth email.", "text": "Exact match of the email to the ground-truth email.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the email to the upper-left or upper-right of the document.", "text": "Location match of the email to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the job/position title to the ground-truth job/position title.", "text": "Exact match of the job/position title to the ground-truth job/position title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the job/position title to the upper-left or upper-right of the document.", "text": "Location match of the job/position title to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the institution/company to the ground-truth institution/company.", "text": "Exact match of the institution/company to the ground-truth institution/company.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the institution/company to the upper-left or upper-right of the document.", "text": "Location match of the institution/company to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The logo was added to the requested place in the document.", "description": "The logo was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an image of the employer's institution logo.", "text": "Image in the doc is an image of the employer's institution logo.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is placed at the top left of the document", "text": "Logo is placed at the top left of the document", "category": "spatial_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The signature was added to the requested place in the document.", "description": "The signature was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an exact match to the ground-truth signature.", "text": "Image in the doc is an exact match to the ground-truth signature.", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "text": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 12, "total_points": 12, "requires_browsing_history": false, "source_assets": ["https://drive.google.com/drive/folders/1ykj346THKjAlKUDUNHHZD0_C84PUASp7"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_4/checkpoints.md"}
{"task_id": "knows.docs_1_formal_letter.5", "short_id": "1e", "family": "docs_1_formal_letter", "instance": "instance_5", "instance_num": 5, "artifact": "Docs", "group": "Letters", "domain": "Business Letter", "prompt": "Structure a formal letter written in my name using Google Docs.\nUse my website for my info: https://mttk.github.io/.\nInclude my employer's logo in the upper left corner if available online. Use the first affiliation listed if multiple exist. For incoming positions, use the future institution.\nInclude the following information in the letter header:\n\n- My Name\n- My Email Address\n- My Github\n- My LinkedIn\n\nTake the photo of my signature from the drive folder here: https://drive.google.com/drive/folders/1yr55MnD387BOSULTrEBOBfOtxtyMJGnm and add it to the end of the letter.", "checkpoints": [{"n": 1, "title": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "description": "Check that common information about the letter writer is placed at a reasonable location in the document letter.", "steps": [{"name": null, "description": "Exact match of the name to the ground-truth name.", "text": "Exact match of the name to the ground-truth name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the name to the upper-left or upper-right of the document.", "text": "Location match of the name to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the email to the ground-truth email.", "text": "Exact match of the email to the ground-truth email.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the email to the upper-left or upper-right of the document.", "text": "Location match of the email to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the Github URL/username to the ground-truth Github.", "text": "Exact match of the Github URL/username to the ground-truth Github.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the Github to the upper-left or upper-right of the document.", "text": "Location match of the Github to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Exact match of the LinkedIn URL/username to the ground-truth LinkedIn.", "text": "Exact match of the LinkedIn URL/username to the ground-truth LinkedIn.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Location match of the LinkedIn to the upper-left or upper-right of the document.", "text": "Location match of the LinkedIn to the upper-left or upper-right of the document.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The logo was added to the requested place in the document.", "description": "The logo was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an image of the employer's institution logo.", "text": "Image in the doc is an image of the employer's institution logo.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is placed at the top left of the document", "text": "Logo is placed at the top left of the document", "category": "spatial_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The signature was added to the requested place in the document.", "description": "The signature was added to the requested place in the document.", "steps": [{"name": null, "description": "Image in the doc is an exact match to the ground-truth signature.", "text": "Image in the doc is an exact match to the ground-truth signature.", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "text": "Signature is placed at the bottom of the document, below the letter body and after any other content.", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 12, "total_points": 12, "requires_browsing_history": false, "source_assets": ["https://drive.google.com/drive/folders/1yr55MnD387BOSULTrEBOBfOtxtyMJGnm"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_1_formal_letter/instance_5/checkpoints.md"}
{"task_id": "knows.docs_31_education_lesson_plan.1", "short_id": "31a", "family": "docs_31_education_lesson_plan", "instance": "instance_1", "instance_num": 1, "artifact": "Docs", "group": "Education", "domain": "Lesson Plan", "prompt": "I am a Middle School educator, and I have a free class to teach a fun topic. My class is on Chemistry. Can you help me make a lesson plan document appropriate for my students? First, come up with 2-3 fun topics in Chemistry that the class might enjoy but is still related and list those as the main bullets. Then, under each of these, find 4 unique websites that have information about the topic and put their URLs as sub-bullets. Then, for each of these websites, list 5 facts as sub-bullets to the url. Finally, list all of the facts below the bullets and color code them by which topic they are about with the highlighter. Use a different color for each topic. Write a summary of the lesson using the facts collected. Finally, go online to find images related to each of the topics. Put them all side-by-side at the bottom of the last page of the document.", "checkpoints": [{"n": 1, "title": "Related Topics", "description": "The document has two to three main level topics that match the given subject and are appropriate for the specified audience.", "steps": [{"name": null, "description": "All top level bullet points are topics related to the given subject for the instance.", "text": "All top level bullet points are topics related to the given subject for the instance.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "text": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are unique (no duplicate topic names).", "text": "All topics are unique (no duplicate topic names).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "text": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "text": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Website url's match", "description": "The document has valid urls's under each topic bullet point.", "steps": [{"name": null, "description": "Ensure that each topic has at least 4 websites listed as sub bullets", "text": "Ensure that each topic has at least 4 websites listed as sub bullets", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "text": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is valid", "text": "Ensure that each website is valid", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure the agent has visited each website", "text": "Ensure the agent has visited each website", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is related to the topic", "text": "Ensure that each website is related to the topic", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Website Facts are valid", "description": "Ensures that facts exist in the website.", "steps": [{"name": null, "description": "Ensure that each website bullet has at least 5 sub bullets for facts", "text": "Ensure that each website bullet has at least 5 sub bullets for facts", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "text": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each fact does exist in the website.", "text": "Ensure that each fact does exist in the website.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Check color coded facts", "description": "Ensure that exact facts are copied into the summary and they are properly color coded by topic.", "steps": [{"name": null, "description": "Ensure that all facts are copied in correctly.", "text": "Ensure that all facts are copied in correctly.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the facts are formatted as bullet points.", "text": "Ensure that the facts are formatted as bullet points.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each subtopic facts are all color coded the same", "text": "Ensure that each subtopic facts are all color coded the same", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "text": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Check Topic Images", "description": "Check that there are correct images for each topic and they are correctley aligned.", "steps": [{"name": null, "description": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "text": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the images are at the bottom of the last page", "text": "Ensure that the images are at the bottom of the last page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that images are aligned side by side", "text": "Check that images are aligned side by side", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "text": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Check lesson summary", "description": "Ensure that lesson summary exists and they are written from the facts.", "steps": [{"name": null, "description": "Ensure that lesson summary is present.", "text": "Ensure that lesson summary is present.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the lesson summary is written using the facts collected.", "text": "Ensure that the lesson summary is written using the facts collected.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 23, "total_points": 145, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_1/checkpoints.md"}
{"task_id": "knows.docs_31_education_lesson_plan.2", "short_id": "31b", "family": "docs_31_education_lesson_plan", "instance": "instance_2", "instance_num": 2, "artifact": "Docs", "group": "Education", "domain": "Lesson Plan", "prompt": "I am a High School educator, and I have a free class to teach a fun topic. My class is on Mereology. Can you help me make a lesson plan document appropriate for my students? First, come up with 2-3 fun topics in Mereology that the class might enjoy but is still related and list those as the main bullets. Then, under each of these, find 4 unique websites that have information about the topic and put their URLs as sub-bullets. Then, for each of these websites, list 5 facts as sub-bullets to the url. Finally, list all of the facts below the bullets and color code them by which topic they are about with the highlighter. Use a different color for each topic. Write a summary of the lesson using the facts collected. Finally, go online to find images related to each of the topics. Put them all side-by-side at the bottom of the last page of the document.", "checkpoints": [{"n": 1, "title": "Related Topics", "description": "The document has two to three main level topics that match the given subject and are appropriate for the specified audience.", "steps": [{"name": null, "description": "All top level bullet points are topics related to the given subject for the instance.", "text": "All top level bullet points are topics related to the given subject for the instance.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "text": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are unique (no duplicate topic names).", "text": "All topics are unique (no duplicate topic names).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "text": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "text": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Website url's match", "description": "The document has valid urls's under each topic bullet point.", "steps": [{"name": null, "description": "Ensure that each topic has at least 4 websites listed as sub bullets", "text": "Ensure that each topic has at least 4 websites listed as sub bullets", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "text": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is valid", "text": "Ensure that each website is valid", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure the agent has visited each website", "text": "Ensure the agent has visited each website", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is related to the topic", "text": "Ensure that each website is related to the topic", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Website Facts are valid", "description": "Ensures that facts exist in the website.", "steps": [{"name": null, "description": "Ensure that each website bullet has at least 5 sub bullets for facts", "text": "Ensure that each website bullet has at least 5 sub bullets for facts", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "text": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each fact does exist in the website.", "text": "Ensure that each fact does exist in the website.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Check color coded facts", "description": "Ensure that exact facts are copied into the summary and they are properly color coded by topic.", "steps": [{"name": null, "description": "Ensure that all facts are copied in correctly.", "text": "Ensure that all facts are copied in correctly.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the facts are formatted as bullet points.", "text": "Ensure that the facts are formatted as bullet points.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each subtopic facts are all color coded the same", "text": "Ensure that each subtopic facts are all color coded the same", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "text": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Check Topic Images", "description": "Check that there are correct images for each topic and they are correctley aligned.", "steps": [{"name": null, "description": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "text": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the images are at the bottom of the last page", "text": "Ensure that the images are at the bottom of the last page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that images are aligned side by side", "text": "Check that images are aligned side by side", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "text": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Check lesson summary", "description": "Ensure that lesson summary exists and they are written from the facts.", "steps": [{"name": null, "description": "Ensure that lesson summary is present.", "text": "Ensure that lesson summary is present.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the lesson summary is written using the facts collected.", "text": "Ensure that the lesson summary is written using the facts collected.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 23, "total_points": 145, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_2/checkpoints.md"}
{"task_id": "knows.docs_31_education_lesson_plan.3", "short_id": "31c", "family": "docs_31_education_lesson_plan", "instance": "instance_3", "instance_num": 3, "artifact": "Docs", "group": "Education", "domain": "Lesson Plan", "prompt": "I am an Elementary School educator, and I have a free class to teach a fun topic. My class is on Deep Learning. Can you help me make a lesson plan document appropriate for my students? First, come up with 2-3 fun topics in Deep Learning that the class might enjoy but is still related and list those as the main bullets. Then, under each of these, find 4 unique websites that have information about the topic and put their URLs as sub-bullets. Then, for each of these websites, list 5 facts as sub-bullets to the url. Finally, list all of the facts below the bullets and color code them by which topic they are about with the highlighter. Use a different color for each topic. Write a summary of the lesson using the facts collected. Finally, go online to find images related to each of the topics. Put them all side-by-side at the bottom of the last page of the document.", "checkpoints": [{"n": 1, "title": "Related Topics", "description": "The document has two to three main level topics that match the given subject and are appropriate for the specified audience.", "steps": [{"name": null, "description": "All top level bullet points are topics related to the given subject for the instance.", "text": "All top level bullet points are topics related to the given subject for the instance.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "text": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are unique (no duplicate topic names).", "text": "All topics are unique (no duplicate topic names).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "text": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "text": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Website url's match", "description": "The document has valid urls's under each topic bullet point.", "steps": [{"name": null, "description": "Ensure that each topic has at least 4 websites listed as sub bullets", "text": "Ensure that each topic has at least 4 websites listed as sub bullets", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "text": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is valid", "text": "Ensure that each website is valid", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure the agent has visited each website", "text": "Ensure the agent has visited each website", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is related to the topic", "text": "Ensure that each website is related to the topic", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Website Facts are valid", "description": "Ensures that facts exist in the website.", "steps": [{"name": null, "description": "Ensure that each website bullet has at least 5 sub bullets for facts", "text": "Ensure that each website bullet has at least 5 sub bullets for facts", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "text": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each fact does exist in the website.", "text": "Ensure that each fact does exist in the website.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Check color coded facts", "description": "Ensure that exact facts are copied into the summary and they are properly color coded by topic.", "steps": [{"name": null, "description": "Ensure that all facts are copied in correctly.", "text": "Ensure that all facts are copied in correctly.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the facts are formatted as bullet points.", "text": "Ensure that the facts are formatted as bullet points.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each subtopic facts are all color coded the same", "text": "Ensure that each subtopic facts are all color coded the same", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "text": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Check Topic Images", "description": "Check that there are correct images for each topic and they are correctley aligned.", "steps": [{"name": null, "description": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "text": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the images are at the bottom of the last page", "text": "Ensure that the images are at the bottom of the last page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that images are aligned side by side", "text": "Check that images are aligned side by side", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "text": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Check lesson summary", "description": "Ensure that lesson summary exists and they are written from the facts.", "steps": [{"name": null, "description": "Ensure that lesson summary is present.", "text": "Ensure that lesson summary is present.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the lesson summary is written using the facts collected.", "text": "Ensure that the lesson summary is written using the facts collected.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 23, "total_points": 145, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_3/checkpoints.md"}
{"task_id": "knows.docs_31_education_lesson_plan.4", "short_id": "31d", "family": "docs_31_education_lesson_plan", "instance": "instance_4", "instance_num": 4, "artifact": "Docs", "group": "Education", "domain": "Lesson Plan", "prompt": "I am a College educator, and I have a free class to teach a fun topic. My class is on Statistics. Can you help me make a lesson plan document appropriate for my students? First, come up with 2-3 fun topics in Statistics that the class might enjoy but is still related and list those as the main bullets. Then, under each of these, find 4 unique websites that have information about the topic and put their URLs as sub-bullets. Then, for each of these websites, list 5 facts as sub-bullets to the url. Finally, list all of the facts below the bullets and color code them by which topic they are about with the highlighter. Use a different color for each topic. Write a summary of the lesson using the facts collected. Finally, go online to find images related to each of the topics. Put them all side-by-side at the bottom of the last page of the document.", "checkpoints": [{"n": 1, "title": "Related Topics", "description": "The document has two to three main level topics that match the given subject and are appropriate for the specified audience.", "steps": [{"name": null, "description": "All top level bullet points are topics related to the given subject for the instance.", "text": "All top level bullet points are topics related to the given subject for the instance.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "text": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are unique (no duplicate topic names).", "text": "All topics are unique (no duplicate topic names).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "text": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "text": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Website url's match", "description": "The document has valid urls's under each topic bullet point.", "steps": [{"name": null, "description": "Ensure that each topic has at least 4 websites listed as sub bullets", "text": "Ensure that each topic has at least 4 websites listed as sub bullets", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "text": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is valid", "text": "Ensure that each website is valid", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure the agent has visited each website", "text": "Ensure the agent has visited each website", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is related to the topic", "text": "Ensure that each website is related to the topic", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Website Facts are valid", "description": "Ensures that facts exist in the website.", "steps": [{"name": null, "description": "Ensure that each website bullet has at least 5 sub bullets for facts", "text": "Ensure that each website bullet has at least 5 sub bullets for facts", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "text": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each fact does exist in the website.", "text": "Ensure that each fact does exist in the website.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Check color coded facts", "description": "Ensure that exact facts are copied into the summary and they are properly color coded by topic.", "steps": [{"name": null, "description": "Ensure that all facts are copied in correctly.", "text": "Ensure that all facts are copied in correctly.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the facts are formatted as bullet points.", "text": "Ensure that the facts are formatted as bullet points.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each subtopic facts are all color coded the same", "text": "Ensure that each subtopic facts are all color coded the same", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "text": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Check Topic Images", "description": "Check that there are correct images for each topic and they are correctley aligned.", "steps": [{"name": null, "description": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "text": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the images are at the bottom of the last page", "text": "Ensure that the images are at the bottom of the last page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that images are aligned side by side", "text": "Check that images are aligned side by side", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "text": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Check lesson summary", "description": "Ensure that lesson summary exists and they are written from the facts.", "steps": [{"name": null, "description": "Ensure that lesson summary is present.", "text": "Ensure that lesson summary is present.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the lesson summary is written using the facts collected.", "text": "Ensure that the lesson summary is written using the facts collected.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 23, "total_points": 145, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_4/checkpoints.md"}
{"task_id": "knows.docs_31_education_lesson_plan.5", "short_id": "31e", "family": "docs_31_education_lesson_plan", "instance": "instance_5", "instance_num": 5, "artifact": "Docs", "group": "Education", "domain": "Lesson Plan", "prompt": "I am a Middle School educator, and I have a free class to teach a fun topic. My class is on Earth Science. Can you help me make a lesson plan document appropriate for my students? First, come up with 2-3 fun topics in Earth Science that the class might enjoy but is still related and list those as the main bullets. Then, under each of these, find 4 unique websites that have information about the topic and put their URLs as sub-bullets. Then, for each of these websites, list 5 facts as sub-bullets to the url. Finally, list all of the facts below the bullets and color code them by which topic they are about with the highlighter. Use a different color for each topic. Write a summary of the lesson using the facts collected. Finally, go online to find images related to each of the topics. Put them all side-by-side at the bottom of the last page of the document.", "checkpoints": [{"n": 1, "title": "Related Topics", "description": "The document has two to three main level topics that match the given subject and are appropriate for the specified audience.", "steps": [{"name": null, "description": "All top level bullet points are topics related to the given subject for the instance.", "text": "All top level bullet points are topics related to the given subject for the instance.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "text": "There is at least two main level bullet point topics and no more then three main level bullet point topics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are unique (no duplicate topic names).", "text": "All topics are unique (no duplicate topic names).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "text": "All topics are appropriate for the specified educational level (e.g., not too advanced for elementary, not trivially shallow for college).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "text": "All topics are engaging and likely to be enjoyed by students at the specified level (the \"fun\" criterion from the task).", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Website url's match", "description": "The document has valid urls's under each topic bullet point.", "steps": [{"name": null, "description": "Ensure that each topic has at least 4 websites listed as sub bullets", "text": "Ensure that each topic has at least 4 websites listed as sub bullets", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "text": "Ensure that the 4 websites under each topic are unique (no duplicate URLs within a topic, after normalization)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is valid", "text": "Ensure that each website is valid", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure the agent has visited each website", "text": "Ensure the agent has visited each website", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each website is related to the topic", "text": "Ensure that each website is related to the topic", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Website Facts are valid", "description": "Ensures that facts exist in the website.", "steps": [{"name": null, "description": "Ensure that each website bullet has at least 5 sub bullets for facts", "text": "Ensure that each website bullet has at least 5 sub bullets for facts", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "text": "Ensure that the 5 facts under each website are unique (no duplicate fact text)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each fact does exist in the website.", "text": "Ensure that each fact does exist in the website.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Check color coded facts", "description": "Ensure that exact facts are copied into the summary and they are properly color coded by topic.", "steps": [{"name": null, "description": "Ensure that all facts are copied in correctly.", "text": "Ensure that all facts are copied in correctly.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that the facts are formatted as bullet points.", "text": "Ensure that the facts are formatted as bullet points.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each subtopic facts are all color coded the same", "text": "Ensure that each subtopic facts are all color coded the same", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "text": "Ensure that each topic uses a different color from every other topic (no color shared between topics)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Check Topic Images", "description": "Check that there are correct images for each topic and they are correctley aligned.", "steps": [{"name": null, "description": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "text": "Ensure every topic from the lesson plan has exactly one matching image (one-to-one; a single image must not satisfy multiple topics).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the images are at the bottom of the last page", "text": "Ensure that the images are at the bottom of the last page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that images are aligned side by side", "text": "Check that images are aligned side by side", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "text": "Ensure that each image depicts the topic concept or related people, not merely a surface keyword association.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Check lesson summary", "description": "Ensure that lesson summary exists and they are written from the facts.", "steps": [{"name": null, "description": "Ensure that lesson summary is present.", "text": "Ensure that lesson summary is present.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Ensure that the lesson summary is written using the facts collected.", "text": "Ensure that the lesson summary is written using the facts collected.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 23, "total_points": 145, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_31_education_lesson_plan/instance_5/checkpoints.md"}
{"task_id": "knows.docs_37_reference_list.1", "short_id": "37a", "family": "docs_37_reference_list", "instance": "instance_1", "instance_num": 1, "artifact": "Docs", "group": "Education", "domain": "Class Notes / Reference List", "prompt": "Create a Google Doc reference list from this course webpage: https://utah-cs6340-nlp-f24.notion.site/?pvs=74. For each lecture, add a Heading 3 with 'Month/Day: Lecture title' and create a bookmark. Things like exams and holidays should not appear in the reference list. Only consider Guest Lectures if it has slides, otherwise it should not appear in the reference list. Under each title, create a separate bullet list for academic articles, textbooks, blogs, code implementations, tutorials, wikipedia and presentations. Each bullet list should start with \"Academic Articles\", \"Textbooks\", or \"Blogs\", etc. in bold, normal font. If a given type has no hyperlinks, delete that category. Each bullet is a hyperlink to a resource referenced in the lecture slides, using the resource title as the anchor text instead of the raw URL. Next to the hyperlink, add \"Slide: <list of slide numbers separated by commas>\" listing the slides where the link is referenced. No duplicates: if something is referenced multiple times, then add all slide numbers. If a link is referenced on more than one slide, bold that hyperlink (and only the hyperlink) in the doc and color it 'dark green 2'. If the link is inactive (dead link) do not include it in the google doc.", "checkpoints": [{"n": 1, "title": "Lecture Title", "description": "", "steps": [{"name": null, "description": "Each lecture (and only lectures) has an associated title in 'Month/Day: Lecture title' format.(10 pt, proportional)", "text": "Each lecture (and only lectures) has an associated title in 'Month/Day: Lecture title' format.(10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The title of each lecture should be of size 'Heading 3'(10 pt, proportional)", "text": "The title of each lecture should be of size 'Heading 3'(10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each title should be bookmarked(10 pt, proportional)", "text": "Each title should be bookmarked(10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Topics bullet lists", "description": "", "steps": [{"name": null, "description": "The following categories found as the titles of the bullet lists: \"Academic Articles\", \"Textbooks\", \"Blogs\", \"Code Implementations\", \"Tutorials\", \"Wikipedia\", \"Presentations\".(10 pt, proportional)", "text": "The following categories found as the titles of the bullet lists: \"Academic Articles\", \"Textbooks\", \"Blogs\", \"Code Implementations\", \"Tutorials\", \"Wikipedia\", \"Presentations\".(10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each bullet list starts with the right title in bold.(10 pt, proportional)", "text": "Each bullet list starts with the right title in bold.(10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "There are no empty bullet lists.(10 pt, proportional)", "text": "There are no empty bullet lists.(10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There are no bullet lists with titles that don't match the categories(10 pt, proportional)", "text": "There are no bullet lists with titles that don't match the categories(10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Reference links", "description": "", "steps": [{"name": null, "description": "All reference links under each category of each lecture are present. (10 pt, proportional)", "text": "All reference links under each category of each lecture are present. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "No duplicate reference links. (10 pt, proportional)", "text": "No duplicate reference links. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each link is categorized correctly i.e. a link that is a blog should be categorized as a blog and not as textbooks.(10 pt, proportional)", "text": "Each link is categorized correctly i.e. a link that is a blog should be categorized as a blog and not as textbooks.(10 pt, proportional)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "text": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "text": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "text": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The name of each link is relevant to the webpage it opens.(10 pt, proportional)", "text": "The name of each link is relevant to the webpage it opens.(10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "No dead hyperlink is included in the docs (10pt, proportional)", "text": "No dead hyperlink is included in the docs (10pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Multiple references", "description": "", "steps": [{"name": null, "description": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text).(10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text).(10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are colored 'dark green 2', and only the hyperlink text is colored (not the surrounding bullet text).(10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are colored 'dark green 2', and only the hyperlink text is colored (not the surrounding bullet text).(10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "There are no duplicates in the slide numbers for hyperlinks that appear in multiple slides.(10 pt, proportional)", "text": "There are no duplicates in the slide numbers for hyperlinks that appear in multiple slides.(10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 18, "total_points": 180, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_1/checkpoints.md"}
{"task_id": "knows.docs_37_reference_list.2", "short_id": null, "family": "docs_37_reference_list", "instance": "instance_2", "instance_num": 2, "artifact": "Docs", "group": "Education", "domain": "Class Notes / Reference List", "prompt": "Create a Google Doc reference list from this course webpage: https://sparklabutah.github.io/Multimodal-LLM-Agents-Course-Public/schedule/. For each lecture, add a Heading 3 with 'Module N: Lecture title', where N is the module number. Things like exams and holidays should not appear in the reference list. Under each title, create a separate bullet list for tutorials, textbooks, and videos. Each bullet list should start with \"Tutorials\", \"Textbooks\", or \"Videos\" in bold, normal font. If a given type has no hyperlinks, delete that category. Each bullet is a hyperlink to a resource referenced in the lecture slides, using the resource title as the anchor text instead of the raw URL. Include author(s) in parentheses after the hyperlink, formatted as \"(First author)\". Next to the hyperlink, add \"Slide: <list of slide numbers separated by commas>\" listing the slides where the link is referenced. No duplicates: if something is referenced multiple times, then add all slide numbers. If a link is referenced on more than one slide, bold that hyperlink (and only the hyperlink) in the corresponding bullet list, italicize it, color it 'dark cyan 1', and set its font size to 12pt. If the link is inactive (dead link) do not include it in the google doc.", "checkpoints": [{"n": 1, "title": "Lecture Title", "description": "", "steps": [{"name": null, "description": "Each lecture (and only lectures) has an associated title in 'Module N: Lecture title' format, where N is the module number. (10 pt, proportional)", "text": "Each lecture (and only lectures) has an associated title in 'Module N: Lecture title' format, where N is the module number. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "text": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Topics bullet lists", "description": "", "steps": [{"name": null, "description": "The following categories found as the titles of the bullet lists: \"Tutorials\", \"Textbooks\", \"Videos\". (10 pt, proportional)", "text": "The following categories found as the titles of the bullet lists: \"Tutorials\", \"Textbooks\", \"Videos\". (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "text": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No empty bullet lists. (10 pt, proportional)", "text": "No empty bullet lists. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "text": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Reference links", "description": "", "steps": [{"name": null, "description": "All reference links under each category of each lecture are present. (10 pt, proportional)", "text": "All reference links under each category of each lecture are present. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "No duplicate reference links. (10 pt, proportional)", "text": "No duplicate reference links. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each link is categorized correctly i.e. a link that is a tutorial should be categorized as \"Tutorials\" and not as \"Textbooks\". (10 pt, proportional)", "text": "Each link is categorized correctly i.e. a link that is a tutorial should be categorized as \"Tutorials\" and not as \"Textbooks\". (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "text": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "text": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "text": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Author(s) are included after the hyperlink in the correct format \"(First author)\". (10 pt, proportional)", "text": "Author(s) are included after the hyperlink in the correct format \"(First author)\". (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "text": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "No dead hyperlink is included in the doc. (10 pt, proportional)", "text": "No dead hyperlink is included in the doc. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Multiple references", "description": "", "steps": [{"name": null, "description": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are italicized, and only the hyperlink text is italicized (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are italicized, and only the hyperlink text is italicized (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are colored 'dark cyan 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are colored 'dark cyan 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are set to 12pt font size, and only the hyperlink text is 12pt (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are set to 12pt font size, and only the hyperlink text is 12pt (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "text": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 200, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_2/checkpoints.md"}
{"task_id": "knows.docs_37_reference_list.3", "short_id": null, "family": "docs_37_reference_list", "instance": "instance_3", "instance_num": 3, "artifact": "Docs", "group": "Education", "domain": "Class Notes / Reference List", "prompt": "Create a Google Doc reference list from this course webpage: https://mit6874.github.io/. For each lecture, add a Heading 3 with 'Month/Day: Lecture title'. Things like exams and holidays should not appear in the reference list. Under each title, create a separate bullet list for presentation and blog. Each bullet list should start with \"Presentations\" or \"Blogs\" in bold, normal font. If a given type has no hyperlinks, delete that category. Within each category, sort the bullet points in descending order by publication year, then by title (dead links can be ignored when sorting). Each bullet is a hyperlink to a resource referenced in the lecture slides, using the resource title as the anchor text instead of the raw URL. Include the publication year and first author in parentheses after the hyperlink, formatted as \"(YYYY, First Author)\". Next to the hyperlink, add \"Slide: <list of slide numbers separated by commas>\" listing the slides where the link is referenced. No duplicates: if something is referenced multiple times, then add all slide numbers. If a link is referenced on more than one slide, increase the font size of the hyperlink (and only the hyperlink) by 1pt in the corresponding bullet list, bold it, and color it 'light magenta 1'. If the link is inactive (dead link) include it with strikethrough formatting applied to the anchor text.", "checkpoints": [{"n": 1, "title": "Lecture Title", "description": "", "steps": [{"name": null, "description": "Each lecture (and only lectures) has an associated title in 'Month/Day: Lecture title' format. (10 pt, proportional)", "text": "Each lecture (and only lectures) has an associated title in 'Month/Day: Lecture title' format. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "text": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Topics bullet lists", "description": "", "steps": [{"name": null, "description": "The following categories found as the titles of the bullet lists: \"Presentations\", \"Blogs\". (10 pt, proportional)", "text": "The following categories found as the titles of the bullet lists: \"Presentations\", \"Blogs\". (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "text": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No empty bullet lists. (10 pt, proportional)", "text": "No empty bullet lists. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "text": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Reference links", "description": "", "steps": [{"name": null, "description": "All reference links under each category of each lecture are present. (10 pt, proportional)", "text": "All reference links under each category of each lecture are present. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "No duplicate reference links. (10 pt, proportional)", "text": "No duplicate reference links. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each link is categorized correctly i.e. a link that is a blog should be categorized as \"Blogs\" and not as \"Presentations\". Skipped for dead links. (10 pt, proportional)", "text": "Each link is categorized correctly i.e. a link that is a blog should be categorized as \"Blogs\" and not as \"Presentations\". Skipped for dead links. (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "text": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. Skipped for dead links. (10 pt, proportional)", "text": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. Skipped for dead links. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "text": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Publication year and first author are included after the hyperlink in the correct format \"(YYYY, First Author)\". Skipped for dead links. (10 pt, proportional)", "text": "Publication year and first author are included after the hyperlink in the correct format \"(YYYY, First Author)\". Skipped for dead links. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The name of each link is relevant to the webpage it opens. Skipped for dead links. (10 pt, proportional)", "text": "The name of each link is relevant to the webpage it opens. Skipped for dead links. (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Dead links are present in the document with strikethrough formatting applied to the anchor text. (10 pt, proportional)", "text": "Dead links are present in the document with strikethrough formatting applied to the anchor text. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Within each category, the bullets are sorted in descending order by publication year, then alphabetically by title. Dead links are ignored when checking sort order. (10 pt, proportional)", "text": "Within each category, the bullets are sorted in descending order by publication year, then alphabetically by title. Dead links are ignored when checking sort order. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Multiple references", "description": "", "steps": [{"name": null, "description": "Hyperlinks of references that appear in multiple slides have their font size increased by 1pt, and only the hyperlink text has the increased font size (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides have their font size increased by 1pt, and only the hyperlink text has the increased font size (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are colored 'light magenta 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are colored 'light magenta 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "text": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 200, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_3/checkpoints.md"}
{"task_id": "knows.docs_37_reference_list.4", "short_id": null, "family": "docs_37_reference_list", "instance": "instance_4", "instance_num": 4, "artifact": "Docs", "group": "Education", "domain": "Class Notes / Reference List", "prompt": "Create a Google Doc reference list from this course webpage: https://ocw.mit.edu/courses/12-001-introduction-to-geology-fall-2013/pages/lecture-notes-and-slides/. For each lecture, add a Heading 3 with 'Session #: Lecture title', where '#' is the number of the session. Things like exams and holidays should not appear in the reference list. Under each title, create a separate bullet list for research papers, wikipedia, GitHub repositories, and videos. Each bullet list should start with \"Research Papers\", \"Wikipedia\", \"GitHub Repositories\", or \"Videos\" in bold, normal font. If a given type has no hyperlinks, delete that category. Within each category, sort the bullets in reversed alphabetical order by anchor text, then alphabetically by author's last name (when applicable). Each bullet is a hyperlink to a resource referenced in the lecture slides, using the resource title as the anchor text instead of the raw URL. Next to the hyperlink, add \"Slide: <list of slide numbers separated by commas>\" listing the slides where the link is referenced. No duplicates: if something is referenced multiple times, then add all slide numbers. If a link is referenced on more than one slide, bold that hyperlink (and only the hyperlink) in the corresponding bullet list, remove its underline, and color it 'dark yellow 1'. If the link is inactive (dead link) do not include it in the google doc.", "checkpoints": [{"n": 1, "title": "Lecture Title", "description": "", "steps": [{"name": null, "description": "Each lecture (and only lectures) has an associated title in 'Session #: Lecture title' format, where # is the session number. (10 pt, proportional)", "text": "Each lecture (and only lectures) has an associated title in 'Session #: Lecture title' format, where # is the session number. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "text": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Topics bullet lists", "description": "", "steps": [{"name": null, "description": "The following categories found as the titles of the bullet lists: \"Research Papers\", \"Wikipedia\", \"GitHub Repositories\", \"Videos\". (10 pt, proportional)", "text": "The following categories found as the titles of the bullet lists: \"Research Papers\", \"Wikipedia\", \"GitHub Repositories\", \"Videos\". (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "text": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No empty bullet lists. (10 pt, proportional)", "text": "No empty bullet lists. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "text": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Reference links", "description": "", "steps": [{"name": null, "description": "All reference links under each category of each lecture are present. (10 pt, proportional)", "text": "All reference links under each category of each lecture are present. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "No duplicate reference links. (10 pt, proportional)", "text": "No duplicate reference links. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each link is categorized correctly i.e. a link that is a Wikipedia article should be categorized as \"Wikipedia\" and not as \"Research Papers\". (10 pt, proportional)", "text": "Each link is categorized correctly i.e. a link that is a Wikipedia article should be categorized as \"Wikipedia\" and not as \"Research Papers\". (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "text": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "text": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "text": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "text": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "No dead hyperlink is included in the doc. (10 pt, proportional)", "text": "No dead hyperlink is included in the doc. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Within each category, the bullets are sorted in reversed alphabetical order by anchor text, then alphabetically by author's last name (when applicable). (10 pt, proportional)", "text": "Within each category, the bullets are sorted in reversed alphabetical order by anchor text, then alphabetically by author's last name (when applicable). (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Multiple references", "description": "", "steps": [{"name": null, "description": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides have their underline removed (the hyperlink text is not underlined). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides have their underline removed (the hyperlink text is not underlined). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are colored 'dark yellow 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are colored 'dark yellow 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "text": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 190, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_4/checkpoints.md"}
{"task_id": "knows.docs_37_reference_list.5", "short_id": null, "family": "docs_37_reference_list", "instance": "instance_5", "instance_num": 5, "artifact": "Docs", "group": "Education", "domain": "Class Notes / Reference List", "prompt": "Create a Google Doc reference list from this course webpage: https://ocw.mit.edu/courses/21h-132-the-ancient-world-rome-spring-2017/pages/lecture-notes. For each lecture, add a Heading 3 with 'Session #: Lecture title', where '#' is the number of the session. Things like exams and holidays should not appear in the reference list. Under each title, create a separate bullet list for research paper, wikipedia, and presentation. Each bullet list should start with \"Research Papers\", \"Wikipedia\", or \"Presentations\" in bold, normal font. If a given type has no hyperlinks, delete that category. Within each category, sort the bullets by descending number of slide references, breaking ties alphabetically by anchor text. Each bullet is a hyperlink to a resource referenced in the lecture slides, using the resource title as the anchor text instead of the raw URL. Next to the hyperlink, add \"Slide: <list of slide numbers separated by commas>\" listing the slides where the link is referenced. No duplicates: if something is referenced multiple times, then add all slide numbers. If a link is referenced on more than one slide, bold that hyperlink (and only the hyperlink) in the corresponding bullet list, italicize it, and color it 'dark purple 1'. If the link is inactive (dead link) place it as a bullet point in a separate \"Inactive Links\" section at the end of the document instead of under its lecture heading.", "checkpoints": [{"n": 1, "title": "Lecture Title", "description": "", "steps": [{"name": null, "description": "Each lecture (and only lectures) has an associated title in 'Session #: Lecture title' format, where # is the session number. (10 pt, proportional)", "text": "Each lecture (and only lectures) has an associated title in 'Session #: Lecture title' format, where # is the session number. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "text": "The title of each lecture should be of size 'Heading 3'. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Topics bullet lists", "description": "", "steps": [{"name": null, "description": "The following categories found as the titles of the bullet lists: \"Research Papers\", \"Wikipedia\", \"Presentations\". (10 pt, proportional)", "text": "The following categories found as the titles of the bullet lists: \"Research Papers\", \"Wikipedia\", \"Presentations\". (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "text": "Each bullet list starts with the right title in bold. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No empty bullet lists. (10 pt, proportional)", "text": "No empty bullet lists. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "text": "No bullet lists with titles that don't match the categories. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Reference links", "description": "", "steps": [{"name": null, "description": "All reference links under each category of each lecture are present. (10 pt, proportional)", "text": "All reference links under each category of each lecture are present. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "No duplicate reference links. (10 pt, proportional)", "text": "No duplicate reference links. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each link is categorized correctly i.e. a link that is a Wikipedia article should be categorized as \"Wikipedia\" and not as \"Research Papers\". (10 pt, proportional)", "text": "Each link is categorized correctly i.e. a link that is a Wikipedia article should be categorized as \"Wikipedia\" and not as \"Research Papers\". (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "text": "The slide number list for each reference matches the gold list of slides where the link is actually referenced. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "text": "Each hyperlink is in the correct format i.e. using the resource title as the anchor text instead of the raw URL. (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "text": "The slide numbers are in the correct format of \"Slide: {number1, number2,...}\". (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "text": "The name of each link is relevant to the webpage it opens. (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "A separate \"Inactive Links\" section exists at the end of the document, and every link placed under it is verified to be a dead link. An empty section passes; otherwise each verified dead link earns a point, scaled to 10. (10 pt, proportional)", "text": "A separate \"Inactive Links\" section exists at the end of the document, and every link placed under it is verified to be a dead link. An empty section passes; otherwise each verified dead link earns a point, scaled to 10. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Within each category, the bullets are sorted in descending order by number of slide references, breaking ties alphabetically by anchor text. (10 pt, proportional)", "text": "Within each category, the bullets are sorted in descending order by number of slide references, breaking ties alphabetically by anchor text. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Multiple references", "description": "", "steps": [{"name": null, "description": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are in bold, and only the hyperlink text is bold (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are italicized, and only the hyperlink text is italicized (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are italicized, and only the hyperlink text is italicized (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Hyperlinks of references that appear in multiple slides are colored 'dark purple 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "text": "Hyperlinks of references that appear in multiple slides are colored 'dark purple 1', and only the hyperlink text is colored (not the surrounding bullet text). (10 pt, proportional)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "text": "No duplicates in the slide numbers for hyperlinks that appear in multiple slides. (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 190, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_37_reference_list/instance_5/checkpoints.md"}
{"task_id": "knows.docs_5_influential_papers.1", "short_id": "5a", "family": "docs_5_influential_papers", "instance": "instance_1", "instance_num": 1, "artifact": "Docs", "group": "Education", "domain": "Research Report", "prompt": "Create a google docs with 5 influential papers on high quality dataset creation for Large Language Models. Include links to arxiv.org for each of the papers. The papers should have at least 50 citations and have been first published within the last 3 years, and must have an active page on semanticscholar.org. Include the name of the paper, a link below the name, and the abstract of each of the papers below the link.", "checkpoints": [{"n": 1, "title": "5 different papers were looked at and placed into the google docs.", "description": "5 different papers were looked at and placed into the google docs.", "steps": [{"name": null, "description": "Paper website was accessed at least 5 times for different papers.", "text": "Paper website was accessed at least 5 times for different papers.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Links in the google docs lines up with the pages that were accessed.", "text": "Links in the google docs lines up with the pages that were accessed.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The papers meet the requirements for citation counts and recency.", "description": "The papers meet the requirements for citation counts and recency.", "steps": [{"name": null, "description": "Each paper has at least 50 citations (1pt each).", "text": "Each paper has at least 50 citations (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper is from the last 3 years (1pt each).", "text": "Each paper is from the last 3 years (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "The doc structure for each paper is correct.", "description": "The doc structure for each paper is correct.", "steps": [{"name": null, "description": "Each paper abstract is included in the google docs.", "text": "Each paper abstract is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper title is included in the google docs.", "text": "Each paper title is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper link is included in the google docs.", "text": "Each paper link is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper structure is correct: Title -> Link -> Abstract.", "text": "Each paper structure is correct: Title -> Link -> Abstract.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "The papers are from the correct relevant domain.", "description": "The papers are from the correct relevant domain.", "steps": [{"name": null, "description": "LLM as Judge for the relevance of each paper abstract to high quality dataset creation for Large", "text": "LLM as Judge for the relevance of each paper abstract to high quality dataset creation for Large", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 4, "n_steps": 9, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_1/checkpoints.md"}
{"task_id": "knows.docs_5_influential_papers.2", "short_id": "5b", "family": "docs_5_influential_papers", "instance": "instance_2", "instance_num": 2, "artifact": "Docs", "group": "Education", "domain": "Research Report", "prompt": "Create a Google Docs with five influential papers on parameter-efficient fine-tuning (PEFT). Include links to arxiv.org for each of the papers. The papers should have at least 100 citations, have been first published within the last three years, and must have an active page on semanticscholar.org. Include the name of the paper, a link below the name, and the abstract of each of the papers below the link.", "checkpoints": [{"n": 1, "title": "5 different papers were looked at and placed into the google docs.", "description": "5 different papers were looked at and placed into the google docs.", "steps": [{"name": null, "description": "Paper website was accessed at least 5 times for different papers.", "text": "Paper website was accessed at least 5 times for different papers.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Links in the google docs lines up with the pages that were accessed.", "text": "Links in the google docs lines up with the pages that were accessed.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The papers meet the requirements for citation counts and recency.", "description": "The papers meet the requirements for citation counts and recency.", "steps": [{"name": null, "description": "Each paper has at least 100 citations (1pt each).", "text": "Each paper has at least 100 citations (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper is from the last 3 years (1pt each).", "text": "Each paper is from the last 3 years (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "The doc structure for each paper is correct.", "description": "The doc structure for each paper is correct.", "steps": [{"name": null, "description": "Each paper abstract is included in the google docs.", "text": "Each paper abstract is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper title is included in the google docs.", "text": "Each paper title is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper link is included in the google docs.", "text": "Each paper link is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper structure is correct: Title -> Link -> Abstract.", "text": "Each paper structure is correct: Title -> Link -> Abstract.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "The papers are from the correct relevant domain.", "description": "The papers are from the correct relevant domain.", "steps": [{"name": null, "description": "LLM as Judge for the relevance of each paper abstract to parameter-efficient fine-tuning (PEFT).", "text": "LLM as Judge for the relevance of each paper abstract to parameter-efficient fine-tuning (PEFT).", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 4, "n_steps": 9, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_2/checkpoints.md"}
{"task_id": "knows.docs_5_influential_papers.3", "short_id": "5c", "family": "docs_5_influential_papers", "instance": "instance_3", "instance_num": 3, "artifact": "Docs", "group": "Education", "domain": "Research Report", "prompt": "Create a Google Docs with five influential papers on quantum error correction (QEC). Include links to arxiv.org for each of the papers. The papers should have at least 20 citations, have been first published within the last five years, and must have an active page on semanticscholar.org. Include the name of the paper, a link below the name, and the abstract of each of the papers below the link.", "checkpoints": [{"n": 1, "title": "5 different papers were looked at and placed into the google docs.", "description": "5 different papers were looked at and placed into the google docs.", "steps": [{"name": null, "description": "Paper website was accessed at least 5 times for different papers.", "text": "Paper website was accessed at least 5 times for different papers.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Links in the google docs lines up with the pages that were accessed.", "text": "Links in the google docs lines up with the pages that were accessed.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The papers meet the requirements for citation counts and recency.", "description": "The papers meet the requirements for citation counts and recency.", "steps": [{"name": null, "description": "Each paper has at least 20 citations (1pt each).", "text": "Each paper has at least 20 citations (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper is from the last 5 years (1pt each).", "text": "Each paper is from the last 5 years (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "The doc structure for each paper is correct.", "description": "The doc structure for each paper is correct.", "steps": [{"name": null, "description": "Each paper abstract is included in the google docs.", "text": "Each paper abstract is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper title is included in the google docs.", "text": "Each paper title is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper link is included in the google docs.", "text": "Each paper link is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper structure is correct: Title -> Link -> Abstract.", "text": "Each paper structure is correct: Title -> Link -> Abstract.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "The papers are from the correct relevant domain.", "description": "The papers are from the correct relevant domain.", "steps": [{"name": null, "description": "LLM as Judge for the relevance of each paper abstract to quantum error correction (QEC).", "text": "LLM as Judge for the relevance of each paper abstract to quantum error correction (QEC).", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 4, "n_steps": 9, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_3/checkpoints.md"}
{"task_id": "knows.docs_5_influential_papers.4", "short_id": "5d", "family": "docs_5_influential_papers", "instance": "instance_4", "instance_num": 4, "artifact": "Docs", "group": "Education", "domain": "Research Report", "prompt": "Create a Google Docs with five influential papers on De Novo Protein Design. Include links to biorxiv.org or nature.com for each of the papers. The papers should have at least 50 citations, have been first published within the last five years, and must have an active page on semanticscholar.org or pubmed.ncbi.nlm.nih.gov. Include the name of the paper, a link below the name, and the abstract of each of the papers below the link.", "checkpoints": [{"n": 1, "title": "5 different papers were looked at and placed into the google docs.", "description": "5 different papers were looked at and placed into the google docs.", "steps": [{"name": null, "description": "Paper website was accessed at least 5 times for different papers.", "text": "Paper website was accessed at least 5 times for different papers.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Links in the google docs lines up with the pages that were accessed.", "text": "Links in the google docs lines up with the pages that were accessed.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The papers meet the requirements for citation counts and recency.", "description": "The papers meet the requirements for citation counts and recency.", "steps": [{"name": null, "description": "Each paper has at least 50 citations (1pt each).", "text": "Each paper has at least 50 citations (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper is from the last 5 years (1pt each).", "text": "Each paper is from the last 5 years (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "The doc structure for each paper is correct.", "description": "The doc structure for each paper is correct.", "steps": [{"name": null, "description": "Each paper abstract is included in the google docs.", "text": "Each paper abstract is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper title is included in the google docs.", "text": "Each paper title is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper link is included in the google docs.", "text": "Each paper link is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper structure is correct: Title -> Link -> Abstract.", "text": "Each paper structure is correct: Title -> Link -> Abstract.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "The papers are from the correct relevant domain.", "description": "The papers are from the correct relevant domain.", "steps": [{"name": null, "description": "LLM as Judge for the relevance of each paper abstract to De Novo Protein Design.", "text": "LLM as Judge for the relevance of each paper abstract to De Novo Protein Design.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 4, "n_steps": 9, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_4/checkpoints.md"}
{"task_id": "knows.docs_5_influential_papers.5", "short_id": "5e", "family": "docs_5_influential_papers", "instance": "instance_5", "instance_num": 5, "artifact": "Docs", "group": "Education", "domain": "Research Report", "prompt": "Create a Google Docs with five influential papers on Solid-State Battery Electrolytes, specifically focusing on dendritic suppression or interface stability. Include links to chemrxiv.org or arxiv.org (cond-mat section) for each of the papers. The papers should have at least 25 citations, have been first published within the last 3 years, and must have an active page on semanticscholar.org. Include the name of the paper, a link below the name, and the abstract of each of the papers below the link.", "checkpoints": [{"n": 1, "title": "5 different papers were looked at and placed into the google docs.", "description": "5 different papers were looked at and placed into the google docs.", "steps": [{"name": null, "description": "Paper website was accessed at least 5 times for different papers.", "text": "Paper website was accessed at least 5 times for different papers.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Links in the google docs lines up with the pages that were accessed.", "text": "Links in the google docs lines up with the pages that were accessed.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The papers meet the requirements for citation counts and recency.", "description": "The papers meet the requirements for citation counts and recency.", "steps": [{"name": null, "description": "Each paper has at least 25 citations (1pt each).", "text": "Each paper has at least 25 citations (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper is from the last 3 years (1pt each).", "text": "Each paper is from the last 3 years (1pt each).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "The doc structure for each paper is correct.", "description": "The doc structure for each paper is correct.", "steps": [{"name": null, "description": "Each paper abstract is included in the google docs.", "text": "Each paper abstract is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper title is included in the google docs.", "text": "Each paper title is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper link is included in the google docs.", "text": "Each paper link is included in the google docs.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each paper structure is correct: Title -> Link -> Abstract.", "text": "Each paper structure is correct: Title -> Link -> Abstract.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "The papers are from the correct relevant domain.", "description": "The papers are from the correct relevant domain.", "steps": [{"name": null, "description": "LLM as Judge for the relevance of each paper abstract to Solid-State Battery Electrolytes, specifically dendritic suppression or interface stability.", "text": "LLM as Judge for the relevance of each paper abstract to Solid-State Battery Electrolytes, specifically dendritic suppression or interface stability.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 4, "n_steps": 9, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/docs_5_influential_papers/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_10_paper_sorting.1", "short_id": "10a", "family": "sheets_10_paper_sorting", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Education", "domain": "Research Library", "prompt": "Create a new google sheets document for all the papers in my drive folder https://drive.google.com/drive/folders/108yXVanQvtgm3X0RxpI-HeR85s7PQNxw?usp=sharing. Make a blank google sheets table and for each paper, in column A put the title of the paper, in column B put the paper authors, in column C put the abstract, in column D put a link to the arxiv abstract website, in column E put the links to the google drive pdfs. Additionally, search the web for at least three more papers on arxiv from the first authors of each existing paper (or less if they do not have at least 3 authorships) and put the same information. For new papers, download the PDFs and upload them to the pdfs folder in this drive folder: {{OUTPUT_FOLDER_URL}} and link them in column E. Extract or take a screenshot of Figure 1 for each paper, upload the image to the figures folder in this drive folder: {{OUTPUT_FOLDER_URL}} and display it in column F using =IMAGE() with the Drive URL. In column G put a checkbox and check for each paper that is a new paper. Highlight papers that have a related works sections which mention chain-of-thought in yellow, and make sure to group these papers at the top rows of the sheet. Ensure all of the text in the table is visible and not cut off.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet has the correct column headers.", "steps": [{"name": "Column A Header", "description": "Column A is labeled \"Title\" (or similar).", "text": "**Column A Header:** Column A is labeled \"Title\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column B Header", "description": "Column B is labeled \"Authors\" (or similar).", "text": "**Column B Header:** Column B is labeled \"Authors\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column C Header", "description": "Column C is labeled \"Abstract\" (or similar).", "text": "**Column C Header:** Column C is labeled \"Abstract\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column D Header", "description": "Column D is labeled \"arXiv Link\" (or similar).", "text": "**Column D Header:** Column D is labeled \"arXiv Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column E Header", "description": "Column E is labeled \"Drive Link\" (or similar).", "text": "**Column E Header:** Column E is labeled \"Drive Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column F Header", "description": "Column F is labeled \"Figure 1\" (or similar).", "text": "**Column F Header:** Column F is labeled \"Figure 1\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column G Header", "description": "Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "text": "**Column G Header:** Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Original Papers Validation", "description": "The spreadsheet contains correct information for all original papers from the source Drive folder.", "steps": [{"name": "Titles (Column A)", "description": "X/N papers have correct titles (fuzzy match to original PDF titles).", "text": "**Titles (Column A):** X/N papers have correct titles (fuzzy match to original PDF titles).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/N papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/N papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/N papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/N papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/N papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/N papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "text": "**Drive Links (Column E):** X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "X/N papers have unchecked checkboxes (not new papers).", "text": "**New Papers Checkbox (Column G):** X/N papers have unchecked checkboxes (not new papers).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/N original paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/N original paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "New Papers Discovery", "description": "For each first author of the original papers, at least 3 additional papers by that author have been added to the spreadsheet (or fewer if the author has fewer than 3 other papers on arXiv).", "steps": [{"name": "Author Coverage", "description": "X/N original first authors have 3+ new papers added.", "text": "**Author Coverage:** X/N original first authors have 3+ new papers added.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "New Papers Validation", "description": "The spreadsheet contains correct information for all new papers added by the agent.", "steps": [{"name": "Titles (Column A)", "description": "X/M new papers have correct titles (fuzzy match).", "text": "**Titles (Column A):** X/M new papers have correct titles (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/M new papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/M new papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/M new papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/M new papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/M new papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/M new papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/M new papers have valid Drive URLs in the correct folder.", "text": "**Drive Links (Column E):** X/M new papers have valid Drive URLs in the correct folder.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "text": "**New Papers Checkbox (Column G):** Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/M new paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/M new paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Formatting & Organization", "description": "Rows are correctly highlighted and organized by color grouping.", "steps": [{"name": "Yellow Highlighting", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined. Papers where content couldn't be fetched are excluded from evaluation.", "text": "**Yellow Highlighting:** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined. Papers where content couldn't be fetched are excluded from evaluation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Row Grouping", "description": "Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "text": "**Row Grouping:** Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Text Overflow", "description": "No text in any cell is hidden due to cell overflow issues.", "text": "**Text Overflow:** No text in any cell is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 10, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/108yXVanQvtgm3X0RxpI-HeR85s7PQNxw?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_10_paper_sorting.2", "short_id": "10b", "family": "sheets_10_paper_sorting", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Education", "domain": "Research Library", "prompt": "Create a new google sheets document for all the papers in my drive folder https://drive.google.com/drive/folders/1vJ3XnjNK792DmyA3relBnrIfIKC5NAED?usp=sharing. Make a blank google sheets table and for each paper, in column A put the title of the paper, in column B put the paper authors, in column C put the abstract, in column D put a link to the arxiv abstract website, in column E put the links to the google drive pdfs. Additionally, search the web for at least three more papers on arxiv where the first author of each existing paper is also listed as the first author (or less if they do not have at least 3 first-authorships on arxiv) and put the same information. For new papers, download the PDFs and upload them to the pdfs folder in this drive folder: {{OUTPUT_FOLDER_URL}} and link them in column E. Extract or take a screenshot of Figure 1 for each paper, upload the image to the figures folder in this drive folder: {{OUTPUT_FOLDER_URL}} and display it in column F using =IMAGE() with the Drive URL. In column G put a checkbox and check for each paper that is a new paper. Highlight papers that have a related works sections which mention dark energy in yellow, and make sure to group these papers at the top rows of the sheet. Ensure all of the text in the table is visible and not cut off.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet has the correct column headers.", "steps": [{"name": "Column A Header", "description": "Column A is labeled \"Title\" (or similar).", "text": "**Column A Header:** Column A is labeled \"Title\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column B Header", "description": "Column B is labeled \"Authors\" (or similar).", "text": "**Column B Header:** Column B is labeled \"Authors\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column C Header", "description": "Column C is labeled \"Abstract\" (or similar).", "text": "**Column C Header:** Column C is labeled \"Abstract\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column D Header", "description": "Column D is labeled \"arXiv Link\" (or similar).", "text": "**Column D Header:** Column D is labeled \"arXiv Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column E Header", "description": "Column E is labeled \"Drive Link\" (or similar).", "text": "**Column E Header:** Column E is labeled \"Drive Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column F Header", "description": "Column F is labeled \"Figure 1\" (or similar).", "text": "**Column F Header:** Column F is labeled \"Figure 1\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column G Header", "description": "Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "text": "**Column G Header:** Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Original Papers Validation", "description": "The spreadsheet contains correct information for all original papers from the source Drive folder.", "steps": [{"name": "Titles (Column A)", "description": "X/N papers have correct titles (fuzzy match to original PDF titles).", "text": "**Titles (Column A):** X/N papers have correct titles (fuzzy match to original PDF titles).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/N papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/N papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/N papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/N papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/N papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/N papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "text": "**Drive Links (Column E):** X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "X/N papers have unchecked checkboxes (not new papers).", "text": "**New Papers Checkbox (Column G):** X/N papers have unchecked checkboxes (not new papers).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/N original paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/N original paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "New Papers Discovery", "description": "For each first author of the original papers, at least 3 additional papers where that author is also first author have been added to the spreadsheet (or fewer if the author has fewer than 3 first-authorships on arXiv).", "steps": [{"name": "Author Coverage", "description": "X/N original first authors have 3+ new first-author papers added.", "text": "**Author Coverage:** X/N original first authors have 3+ new first-author papers added.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "New Papers Validation", "description": "The spreadsheet contains correct information for all new papers added by the agent.", "steps": [{"name": "Titles (Column A)", "description": "X/M new papers have correct titles (fuzzy match).", "text": "**Titles (Column A):** X/M new papers have correct titles (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/M new papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/M new papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/M new papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/M new papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/M new papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/M new papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/M new papers have valid Drive URLs in the correct folder.", "text": "**Drive Links (Column E):** X/M new papers have valid Drive URLs in the correct folder.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "text": "**New Papers Checkbox (Column G):** Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/M new paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/M new paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Formatting & Organization", "description": "Rows are correctly highlighted and organized by color grouping.", "steps": [{"name": "Yellow Highlighting", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "text": "**Yellow Highlighting:** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Row Grouping", "description": "Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "text": "**Row Grouping:** Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Text Overflow", "description": "No text in any cell is hidden due to cell overflow issues.", "text": "**Text Overflow:** No text in any cell is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 10, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1vJ3XnjNK792DmyA3relBnrIfIKC5NAED?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_10_paper_sorting.3", "short_id": "10c", "family": "sheets_10_paper_sorting", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Education", "domain": "Research Library", "prompt": "Create a new google sheets document for all the papers in my drive folder https://drive.google.com/drive/folders/1GmdNa8ulDPoshbdBnm-G6nS-VMBjdW7r?usp=sharing. Make a blank google sheets table and for each paper, in column A put the title of the paper, in column B put the paper authors, in column C put the abstract, in column D put a link to the arxiv abstract website, in column E put the links to the google drive pdfs. Additionally, search the web for at least three more papers on arxiv from the top 20 most cited papers of the first authors of each existing paper (or less if they do not have at least 3 authorships on arxiv) and put the same information. For new papers, download the PDFs and upload them to the pdfs folder in this drive folder: {{OUTPUT_FOLDER_URL}} and link them in column E. Extract or take a screenshot of Figure 1 for each paper, upload the image to the figures folder in this drive folder: {{OUTPUT_FOLDER_URL}} and display it in column F using =IMAGE() with the Drive URL. In column G put a checkbox and check for each paper that is a new paper. Highlight papers that have a related works sections which mention Monte Carlo in yellow, and make sure to group these papers at the top rows of the sheet. Ensure all of the text in the table is visible and not cut off.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet has the correct column headers.", "steps": [{"name": "Column A Header", "description": "Column A is labeled \"Title\" (or similar).", "text": "**Column A Header:** Column A is labeled \"Title\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column B Header", "description": "Column B is labeled \"Authors\" (or similar).", "text": "**Column B Header:** Column B is labeled \"Authors\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column C Header", "description": "Column C is labeled \"Abstract\" (or similar).", "text": "**Column C Header:** Column C is labeled \"Abstract\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column D Header", "description": "Column D is labeled \"arXiv Link\" (or similar).", "text": "**Column D Header:** Column D is labeled \"arXiv Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column E Header", "description": "Column E is labeled \"Drive Link\" (or similar).", "text": "**Column E Header:** Column E is labeled \"Drive Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column F Header", "description": "Column F is labeled \"Figure 1\" (or similar).", "text": "**Column F Header:** Column F is labeled \"Figure 1\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column G Header", "description": "Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "text": "**Column G Header:** Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Original Papers Validation", "description": "The spreadsheet contains correct information for all original papers from the source Drive folder.", "steps": [{"name": "Titles (Column A)", "description": "X/N papers have correct titles (fuzzy match to original PDF titles).", "text": "**Titles (Column A):** X/N papers have correct titles (fuzzy match to original PDF titles).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/N papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/N papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/N papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/N papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/N papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/N papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "text": "**Drive Links (Column E):** X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "X/N papers have unchecked checkboxes (not new papers).", "text": "**New Papers Checkbox (Column G):** X/N papers have unchecked checkboxes (not new papers).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/N original paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/N original paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "New Papers Discovery", "description": "For each first author of the original papers, at least 3 additional papers by that author have been added to the spreadsheet (or fewer if the author has fewer than 3 other papers on arXiv).", "steps": [{"name": "Author Coverage", "description": "X/N original first authors have 3+ new papers added.", "text": "**Author Coverage:** X/N original first authors have 3+ new papers added.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "New Papers Validation", "description": "The spreadsheet contains correct information for all new papers added by the agent.", "steps": [{"name": "Titles (Column A)", "description": "X/M new papers have correct titles (fuzzy match).", "text": "**Titles (Column A):** X/M new papers have correct titles (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/M new papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/M new papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/M new papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/M new papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/M new papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/M new papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/M new papers have valid Drive URLs in the correct folder.", "text": "**Drive Links (Column E):** X/M new papers have valid Drive URLs in the correct folder.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "text": "**New Papers Checkbox (Column G):** Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/M new paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/M new paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Formatting & Organization", "description": "Rows are correctly highlighted and organized by color grouping.", "steps": [{"name": "Yellow Highlighting", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "text": "**Yellow Highlighting:** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Row Grouping", "description": "Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "text": "**Row Grouping:** Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Text Overflow", "description": "No text in any cell is hidden due to cell overflow issues.", "text": "**Text Overflow:** No text in any cell is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 10, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1GmdNa8ulDPoshbdBnm-G6nS-VMBjdW7r?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_10_paper_sorting.4", "short_id": "10d", "family": "sheets_10_paper_sorting", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Education", "domain": "Research Library", "prompt": "Create a new google sheets document for all the papers in my drive folder https://drive.google.com/drive/folders/1rwjStnZZkn4PuF_MV1pEokRkbpAKHv_w?usp=sharing. Make a blank google sheets table and for each paper, in column A put the title of the paper, in column B put the paper authors, in column C put the abstract, in column D put a link to the arxiv abstract website, in column E put the links to the google drive pdfs. Additionally, search the web for at least three more papers on arxiv from the first authors of each existing paper (or less if they do not have at least 3 authorships) and put the same information. For new papers, download the PDFs and upload them to the pdfs folder in this drive folder: {{OUTPUT_FOLDER_URL}} and link them in column E. Extract or take a screenshot of Figure 1 for each paper, upload the image to the figures folder in this drive folder: {{OUTPUT_FOLDER_URL}} and display it in column F using =IMAGE() with the Drive URL. In column G put a checkbox and check for each paper that is a new paper. Highlight papers that have a related works sections which mention deep learning in yellow, and make sure to group these papers at the top rows of the sheet. Ensure all of the text in the table is visible and not cut off.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet has the correct column headers.", "steps": [{"name": "Column A Header", "description": "Column A is labeled \"Title\" (or similar).", "text": "**Column A Header:** Column A is labeled \"Title\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column B Header", "description": "Column B is labeled \"Authors\" (or similar).", "text": "**Column B Header:** Column B is labeled \"Authors\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column C Header", "description": "Column C is labeled \"Abstract\" (or similar).", "text": "**Column C Header:** Column C is labeled \"Abstract\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column D Header", "description": "Column D is labeled \"arXiv Link\" (or similar).", "text": "**Column D Header:** Column D is labeled \"arXiv Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column E Header", "description": "Column E is labeled \"Drive Link\" (or similar).", "text": "**Column E Header:** Column E is labeled \"Drive Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column F Header", "description": "Column F is labeled \"Figure 1\" (or similar).", "text": "**Column F Header:** Column F is labeled \"Figure 1\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column G Header", "description": "Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "text": "**Column G Header:** Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Original Papers Validation", "description": "The spreadsheet contains correct information for all original papers from the source Drive folder.", "steps": [{"name": "Titles (Column A)", "description": "X/N papers have correct titles (fuzzy match to original PDF titles).", "text": "**Titles (Column A):** X/N papers have correct titles (fuzzy match to original PDF titles).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/N papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/N papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/N papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/N papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/N papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/N papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "text": "**Drive Links (Column E):** X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "X/N papers have unchecked checkboxes (not new papers).", "text": "**New Papers Checkbox (Column G):** X/N papers have unchecked checkboxes (not new papers).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/N original paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/N original paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "New Papers Discovery", "description": "For each first author of the original papers, at least 3 additional papers by that author have been added to the spreadsheet (or fewer if the author has fewer than 3 other papers on arXiv).", "steps": [{"name": "Author Coverage", "description": "X/N original first authors have 3+ new papers added.", "text": "**Author Coverage:** X/N original first authors have 3+ new papers added.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "New Papers Validation", "description": "The spreadsheet contains correct information for all new papers added by the agent.", "steps": [{"name": "Titles (Column A)", "description": "X/M new papers have correct titles (fuzzy match).", "text": "**Titles (Column A):** X/M new papers have correct titles (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/M new papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/M new papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/M new papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/M new papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/M new papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/M new papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/M new papers have valid Drive URLs in the correct folder.", "text": "**Drive Links (Column E):** X/M new papers have valid Drive URLs in the correct folder.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "text": "**New Papers Checkbox (Column G):** Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/M new paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/M new paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Formatting & Organization", "description": "Rows are correctly highlighted and organized by color grouping.", "steps": [{"name": "Yellow Highlighting", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "text": "**Yellow Highlighting:** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Row Grouping", "description": "Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "text": "**Row Grouping:** Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Text Overflow", "description": "No text in any cell is hidden due to cell overflow issues.", "text": "**Text Overflow:** No text in any cell is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 10, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1rwjStnZZkn4PuF_MV1pEokRkbpAKHv_w?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_10_paper_sorting.5", "short_id": "10e", "family": "sheets_10_paper_sorting", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Education", "domain": "Research Library", "prompt": "Create a new google sheets document for all the papers in my drive folder https://drive.google.com/drive/folders/1jEEDLHgFdzfV6DCHdTcE2dGDHTHwC0IL?usp=sharing. Make a blank google sheets table and for each paper, in column A put the title of the paper, in column B put the paper authors, in column C put the abstract, in column D put a link to the arxiv abstract website, in column E put the links to the google drive pdfs. Additionally, search the web for at least three more papers on arxiv from the first authors of each existing paper (or less if they do not have at least 3 authorships) and put the same information. For new papers, download the PDFs and upload them to the pdfs folder in this drive folder: {{OUTPUT_FOLDER_URL}} and link them in column E. Extract or take a screenshot of Figure 1 for each paper, upload the image to the figures folder in this drive folder: {{OUTPUT_FOLDER_URL}} and display it in column F using =IMAGE() with the Drive URL. In column G put a checkbox and check for each paper that is a new paper. Highlight papers that have a related works sections which mention Bayesian in yellow, and make sure to group these papers at the top rows of the sheet. Ensure all of the text in the table is visible and not cut off.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet has the correct column headers.", "steps": [{"name": "Column A Header", "description": "Column A is labeled \"Title\" (or similar).", "text": "**Column A Header:** Column A is labeled \"Title\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column B Header", "description": "Column B is labeled \"Authors\" (or similar).", "text": "**Column B Header:** Column B is labeled \"Authors\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column C Header", "description": "Column C is labeled \"Abstract\" (or similar).", "text": "**Column C Header:** Column C is labeled \"Abstract\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column D Header", "description": "Column D is labeled \"arXiv Link\" (or similar).", "text": "**Column D Header:** Column D is labeled \"arXiv Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column E Header", "description": "Column E is labeled \"Drive Link\" (or similar).", "text": "**Column E Header:** Column E is labeled \"Drive Link\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column F Header", "description": "Column F is labeled \"Figure 1\" (or similar).", "text": "**Column F Header:** Column F is labeled \"Figure 1\" (or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Column G Header", "description": "Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "text": "**Column G Header:** Column G is labeled \"New Papers\" (or similar) and is a column of checkboxes.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Original Papers Validation", "description": "The spreadsheet contains correct information for all original papers from the source Drive folder.", "steps": [{"name": "Titles (Column A)", "description": "X/N papers have correct titles (fuzzy match to original PDF titles).", "text": "**Titles (Column A):** X/N papers have correct titles (fuzzy match to original PDF titles).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/N papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/N papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/N papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/N papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/N papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/N papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "text": "**Drive Links (Column E):** X/N papers have valid Drive URLs pointing to correct PDFs (new papers in pdfs/ subfolder).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10). Figures uploaded to figures/ subfolder and displayed via =IMAGE().", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "X/N papers have unchecked checkboxes (not new papers).", "text": "**New Papers Checkbox (Column G):** X/N papers have unchecked checkboxes (not new papers).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/N original paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/N original paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "New Papers Discovery", "description": "For each first author of the original papers, at least 3 additional papers by that author have been added to the spreadsheet (or fewer if the author has fewer than 3 other papers on arXiv).", "steps": [{"name": "Author Coverage", "description": "X/N original first authors have 3+ new papers added.", "text": "**Author Coverage:** X/N original first authors have 3+ new papers added.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "New Papers Validation", "description": "The spreadsheet contains correct information for all new papers added by the agent.", "steps": [{"name": "Titles (Column A)", "description": "X/M new papers have correct titles (fuzzy match).", "text": "**Titles (Column A):** X/M new papers have correct titles (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Authors (Column B)", "description": "X/M new papers have correct author lists (exact match).", "text": "**Authors (Column B):** X/M new papers have correct author lists (exact match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Abstracts (Column C)", "description": "X/M new papers have correct abstracts (fuzzy match).", "text": "**Abstracts (Column C):** X/M new papers have correct abstracts (fuzzy match).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv Links (Column D)", "description": "X/M new papers have valid arXiv URLs pointing to correct papers.", "text": "**arXiv Links (Column D):** X/M new papers have valid arXiv URLs pointing to correct papers.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Drive Links (Column E)", "description": "X/M new papers have valid Drive URLs in the correct folder.", "text": "**Drive Links (Column E):** X/M new papers have valid Drive URLs in the correct folder.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Figure 1 Images (Column F)", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "text": "**Figure 1 Images (Column F):** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers with extractable gold figures.", "category": "visual_check", "llm_vlm_primary": true}, {"name": "New Papers Checkbox (Column G)", "description": "Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "text": "**New Papers Checkbox (Column G):** Proportional (0-10 pts) — scored as floor(checked/total_new_rows × 10) across all non-original rows.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "arXiv URLs Visited", "description": "X/M new paper arXiv URLs appear in the agent browsing history.", "text": "**arXiv URLs Visited:** X/M new paper arXiv URLs appear in the agent browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Formatting & Organization", "description": "Rows are correctly highlighted and organized by color grouping.", "steps": [{"name": "Yellow Highlighting", "description": "Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "text": "**Yellow Highlighting:** Proportional (0-10 pts) — scored as floor(correct/evaluated × 10) where evaluated = papers whose keyword status was successfully determined.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Row Grouping", "description": "Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "text": "**Row Grouping:** Binary (pass/fail) - rows must be grouped by highlight color (not interleaved).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Text Overflow", "description": "No text in any cell is hidden due to cell overflow issues.", "text": "**Text Overflow:** No text in any cell is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 10, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1jEEDLHgFdzfV6DCHdTcE2dGDHTHwC0IL?usp=sharing."], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_10_paper_sorting/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_25_skitourplan.1", "short_id": "25a", "family": "sheets_25_skitourplan", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Outdoor Activity Planner", "prompt": "Create a ski tour plan for 3 possible runs 26 degrees or less using the wasatch backcountry ski guide. Organize the route information for each run in a google sheet including the run name, run link, starting location, gps coordinates of the run location, typical vertical of the run, slope aspect, and slope angle as columns. Then look up the danger color rating for this run according to the Utah Avalanche Center forecast for 02/05/2025. Add columns for forecast date and forecast link and a screenshot of the full danger rose. These 3 columns should be merged vertically since they apply to all runs. Then using this information, determine the danger color rating for each run. Color the cell of each run name with the color of its danger rating.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns for ski run information and avalanche forecast data.", "steps": [{"name": "Run Name Column", "description": "There is a column for the run name.", "text": "**Run Name Column:** There is a column for the run name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Run Link Column", "description": "There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "text": "**Run Link Column:** There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Starting Location Column", "description": "There is a column for the starting location.", "text": "**Starting Location Column:** There is a column for the starting location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "GPS Coordinates Column", "description": "There is a column for GPS coordinates of the run location.", "text": "**GPS Coordinates Column:** There is a column for GPS coordinates of the run location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Typical Vertical Column", "description": "There is a column for typical vertical of the run.", "text": "**Typical Vertical Column:** There is a column for typical vertical of the run.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Aspect Column", "description": "There is a column for slope aspect.", "text": "**Slope Aspect Column:** There is a column for slope aspect.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Column", "description": "There is a column for slope angle.", "text": "**Slope Angle Column:** There is a column for slope angle.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Date Column", "description": "There is a column for forecast date (merged vertically).", "text": "**Forecast Date Column:** There is a column for forecast date (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Link Column", "description": "There is a column for forecast link (merged vertically).", "text": "**Forecast Link Column:** There is a column for forecast link (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Column", "description": "There is a column for the danger rose screenshot (merged vertically).", "text": "**Danger Rose Screenshot Column:** There is a column for the danger rose screenshot (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Run Selection Criteria", "description": "The spreadsheet contains exactly 3 ski runs that meet the slope angle requirement.", "steps": [{"name": "Run Count", "description": "The spreadsheet contains exactly 3 ski runs.", "text": "**Run Count:** The spreadsheet contains exactly 3 ski runs.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 1)", "description": "Run 1 has a slope angle of 26 degrees or less.", "text": "**Slope Angle Compliance (Run 1):** Run 1 has a slope angle of 26 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 2)", "description": "Run 2 has a slope angle of 26 degrees or less.", "text": "**Slope Angle Compliance (Run 2):** Run 2 has a slope angle of 26 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 3)", "description": "Run 3 has a slope angle of 26 degrees or less.", "text": "**Slope Angle Compliance (Run 3):** Run 3 has a slope angle of 26 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Run Data Accuracy", "description": "Each run's data matches information from the Wasatch Backcountry Ski Guide. (7 pts per run x 3 runs)", "steps": [{"name": "Run Name Valid", "description": "The run name exists in the Wasatch Backcountry Ski Guide.", "text": "**Run Name Valid:** The run name exists in the Wasatch Backcountry Ski Guide.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Run Link Valid", "description": "The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "text": "**Run Link Valid:** The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Starting Location Correct", "description": "The starting location matches the guide information.", "text": "**Starting Location Correct:** The starting location matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "GPS Coordinates Correct", "description": "The GPS coordinates match the guide information (within reasonable tolerance).", "text": "**GPS Coordinates Correct:** The GPS coordinates match the guide information (within reasonable tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Typical Vertical Correct", "description": "The typical vertical matches the guide information.", "text": "**Typical Vertical Correct:** The typical vertical matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Aspect Correct", "description": "The slope aspect matches the guide information.", "text": "**Slope Aspect Correct:** The slope aspect matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Correct", "description": "The slope angle matches the guide information and is <= 26 degrees.", "text": "**Slope Angle Correct:** The slope angle matches the guide information and is <= 26 degrees.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather information.", "steps": [{"name": "Wasatch Guide Visited", "description": "The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "text": "**Wasatch Guide Visited:** The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Utah Avalanche Center Visited", "description": "The browsing history contains a visit to the Utah Avalanche Center website.", "text": "**Utah Avalanche Center Visited:** The browsing history contains a visit to the Utah Avalanche Center website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Run Links Visited", "description": "At least one of the run links from the spreadsheet appears in the browsing history.", "text": "**Run Links Visited:** At least one of the run links from the spreadsheet appears in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Forecast Page Visited", "description": "The browsing history contains a visit to the avalanche forecast page for the relevant date.", "text": "**Forecast Page Visited:** The browsing history contains a visit to the avalanche forecast page for the relevant date.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Avalanche Forecast Data", "description": "The avalanche forecast information is correctly included and formatted.", "steps": [{"name": "Forecast Date Correct", "description": "The forecast date is 02/05/2025.", "text": "**Forecast Date Correct:** The forecast date is 02/05/2025.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Forecast Link Valid", "description": "The forecast link is a valid URL to the Utah Avalanche Center forecast.", "text": "**Forecast Link Valid:** The forecast link is a valid URL to the Utah Avalanche Center forecast.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Merged Cells", "description": "The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "text": "**Merged Cells:** The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Present", "description": "An image of the danger rose is present in the spreadsheet.", "text": "**Danger Rose Screenshot Present:** An image of the danger rose is present in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Image Valid", "description": "The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "text": "**Danger Rose Image Valid:** The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Danger Rating Cell Coloring", "description": "Each run name cell is colored according to its avalanche danger rating. (2 pts per run x 3 runs)", "steps": [{"name": "Danger Rating Determined", "description": "The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "text": "**Danger Rating Determined:** The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Cell Color Applied", "description": "The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "text": "**Cell Color Applied:** The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 32, "total_points": 50, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_25_skitourplan.2", "short_id": "25b", "family": "sheets_25_skitourplan", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Outdoor Activity Planner", "prompt": "Create a ski tour plan for 4 possible runs 30 degrees or less using the wasatch backcountry ski guide. Organize the route information for each run in a google sheet including the run name, run link, starting location, gps coordinates of the run location, typical vertical of the run, slope aspect, and slope angle as columns. Then look up the danger color rating for this run according to the Utah Avalanche Center forecast for 01/15/2025. Add columns for forecast date and forecast link and a screenshot of the full danger rose. These 3 columns should be merged vertically since they apply to all runs. Then using this information, determine the danger color rating for each run. Color the cell of each run name with the color of its danger rating.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns for ski run information and avalanche forecast data.", "steps": [{"name": "Run Name Column", "description": "There is a column for the run name.", "text": "**Run Name Column:** There is a column for the run name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Run Link Column", "description": "There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "text": "**Run Link Column:** There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Starting Location Column", "description": "There is a column for the starting location.", "text": "**Starting Location Column:** There is a column for the starting location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "GPS Coordinates Column", "description": "There is a column for GPS coordinates of the run location.", "text": "**GPS Coordinates Column:** There is a column for GPS coordinates of the run location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Typical Vertical Column", "description": "There is a column for typical vertical of the run.", "text": "**Typical Vertical Column:** There is a column for typical vertical of the run.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Aspect Column", "description": "There is a column for slope aspect.", "text": "**Slope Aspect Column:** There is a column for slope aspect.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Column", "description": "There is a column for slope angle.", "text": "**Slope Angle Column:** There is a column for slope angle.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Date Column", "description": "There is a column for forecast date (merged vertically).", "text": "**Forecast Date Column:** There is a column for forecast date (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Link Column", "description": "There is a column for forecast link (merged vertically).", "text": "**Forecast Link Column:** There is a column for forecast link (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Column", "description": "There is a column for the danger rose screenshot (merged vertically).", "text": "**Danger Rose Screenshot Column:** There is a column for the danger rose screenshot (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Run Selection Criteria", "description": "The spreadsheet contains exactly 4 ski runs that meet the slope angle requirement.", "steps": [{"name": "Run Count", "description": "The spreadsheet contains exactly 4 ski runs.", "text": "**Run Count:** The spreadsheet contains exactly 4 ski runs.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 1)", "description": "Run 1 has a slope angle of 30 degrees or less.", "text": "**Slope Angle Compliance (Run 1):** Run 1 has a slope angle of 30 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 2)", "description": "Run 2 has a slope angle of 30 degrees or less.", "text": "**Slope Angle Compliance (Run 2):** Run 2 has a slope angle of 30 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 3)", "description": "Run 3 has a slope angle of 30 degrees or less.", "text": "**Slope Angle Compliance (Run 3):** Run 3 has a slope angle of 30 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 4)", "description": "Run 4 has a slope angle of 30 degrees or less.", "text": "**Slope Angle Compliance (Run 4):** Run 4 has a slope angle of 30 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Run Data Accuracy", "description": "Each run's data matches information from the Wasatch Backcountry Ski Guide. (7 pts per run x 4 runs)", "steps": [{"name": "Run Name Valid", "description": "The run name exists in the Wasatch Backcountry Ski Guide.", "text": "**Run Name Valid:** The run name exists in the Wasatch Backcountry Ski Guide.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Run Link Valid", "description": "The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "text": "**Run Link Valid:** The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Starting Location Correct", "description": "The starting location matches the guide information.", "text": "**Starting Location Correct:** The starting location matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "GPS Coordinates Correct", "description": "The GPS coordinates match the guide information (within reasonable tolerance).", "text": "**GPS Coordinates Correct:** The GPS coordinates match the guide information (within reasonable tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Typical Vertical Correct", "description": "The typical vertical matches the guide information.", "text": "**Typical Vertical Correct:** The typical vertical matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Aspect Correct", "description": "The slope aspect matches the guide information.", "text": "**Slope Aspect Correct:** The slope aspect matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Correct", "description": "The slope angle matches the guide information and is <= 30 degrees.", "text": "**Slope Angle Correct:** The slope angle matches the guide information and is <= 30 degrees.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather information.", "steps": [{"name": "Wasatch Guide Visited", "description": "The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "text": "**Wasatch Guide Visited:** The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Utah Avalanche Center Visited", "description": "The browsing history contains a visit to the Utah Avalanche Center website.", "text": "**Utah Avalanche Center Visited:** The browsing history contains a visit to the Utah Avalanche Center website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Run Links Visited", "description": "At least one of the run links from the spreadsheet appears in the browsing history.", "text": "**Run Links Visited:** At least one of the run links from the spreadsheet appears in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Forecast Page Visited", "description": "The browsing history contains a visit to the avalanche forecast page for the relevant date.", "text": "**Forecast Page Visited:** The browsing history contains a visit to the avalanche forecast page for the relevant date.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Avalanche Forecast Data", "description": "The avalanche forecast information is correctly included and formatted.", "steps": [{"name": "Forecast Date Correct", "description": "The forecast date is 01/15/2025.", "text": "**Forecast Date Correct:** The forecast date is 01/15/2025.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Forecast Link Valid", "description": "The forecast link is a valid URL to the Utah Avalanche Center forecast.", "text": "**Forecast Link Valid:** The forecast link is a valid URL to the Utah Avalanche Center forecast.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Merged Cells", "description": "The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "text": "**Merged Cells:** The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Present", "description": "An image of the danger rose is present in the spreadsheet.", "text": "**Danger Rose Screenshot Present:** An image of the danger rose is present in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Image Valid", "description": "The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "text": "**Danger Rose Image Valid:** The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Danger Rating Cell Coloring", "description": "Each run name cell is colored according to its avalanche danger rating. (2 pts per run x 4 runs, but capped for eval step structure)", "steps": [{"name": "Danger Rating Determined", "description": "The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "text": "**Danger Rating Determined:** The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Cell Color Applied", "description": "The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "text": "**Cell Color Applied:** The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 33, "total_points": 60, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_25_skitourplan.3", "short_id": "25c", "family": "sheets_25_skitourplan", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Outdoor Activity Planner", "prompt": "Create a ski tour plan for 3 possible runs 28 degrees or less using the wasatch backcountry ski guide. Organize the route information for each run in a google sheet including the run name, run link, starting location, gps coordinates of the run location, typical vertical of the run, slope aspect, and slope angle as columns. Then look up the danger color rating for this run according to the Utah Avalanche Center forecast for 02/12/2025. Add columns for forecast date and forecast link and a screenshot of the full danger rose. These 3 columns should be merged vertically since they apply to all runs. Then using this information, determine the danger color rating for each run. Color the cell of each run name with the color of its danger rating.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns for ski run information and avalanche forecast data.", "steps": [{"name": "Run Name Column", "description": "There is a column for the run name.", "text": "**Run Name Column:** There is a column for the run name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Run Link Column", "description": "There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "text": "**Run Link Column:** There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Starting Location Column", "description": "There is a column for the starting location.", "text": "**Starting Location Column:** There is a column for the starting location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "GPS Coordinates Column", "description": "There is a column for GPS coordinates of the run location.", "text": "**GPS Coordinates Column:** There is a column for GPS coordinates of the run location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Typical Vertical Column", "description": "There is a column for typical vertical of the run.", "text": "**Typical Vertical Column:** There is a column for typical vertical of the run.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Aspect Column", "description": "There is a column for slope aspect.", "text": "**Slope Aspect Column:** There is a column for slope aspect.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Column", "description": "There is a column for slope angle.", "text": "**Slope Angle Column:** There is a column for slope angle.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Date Column", "description": "There is a column for forecast date (merged vertically).", "text": "**Forecast Date Column:** There is a column for forecast date (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Link Column", "description": "There is a column for forecast link (merged vertically).", "text": "**Forecast Link Column:** There is a column for forecast link (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Column", "description": "There is a column for the danger rose screenshot (merged vertically).", "text": "**Danger Rose Screenshot Column:** There is a column for the danger rose screenshot (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Run Selection Criteria", "description": "The spreadsheet contains exactly 3 ski runs that meet the slope angle requirement.", "steps": [{"name": "Run Count", "description": "The spreadsheet contains exactly 3 ski runs.", "text": "**Run Count:** The spreadsheet contains exactly 3 ski runs.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 1)", "description": "Run 1 has a slope angle of 28 degrees or less.", "text": "**Slope Angle Compliance (Run 1):** Run 1 has a slope angle of 28 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 2)", "description": "Run 2 has a slope angle of 28 degrees or less.", "text": "**Slope Angle Compliance (Run 2):** Run 2 has a slope angle of 28 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 3)", "description": "Run 3 has a slope angle of 28 degrees or less.", "text": "**Slope Angle Compliance (Run 3):** Run 3 has a slope angle of 28 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Run Data Accuracy", "description": "Each run's data matches information from the Wasatch Backcountry Ski Guide. (7 pts per run x 3 runs)", "steps": [{"name": "Run Name Valid", "description": "The run name exists in the Wasatch Backcountry Ski Guide.", "text": "**Run Name Valid:** The run name exists in the Wasatch Backcountry Ski Guide.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Run Link Valid", "description": "The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "text": "**Run Link Valid:** The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Starting Location Correct", "description": "The starting location matches the guide information.", "text": "**Starting Location Correct:** The starting location matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "GPS Coordinates Correct", "description": "The GPS coordinates match the guide information (within reasonable tolerance).", "text": "**GPS Coordinates Correct:** The GPS coordinates match the guide information (within reasonable tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Typical Vertical Correct", "description": "The typical vertical matches the guide information.", "text": "**Typical Vertical Correct:** The typical vertical matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Aspect Correct", "description": "The slope aspect matches the guide information.", "text": "**Slope Aspect Correct:** The slope aspect matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Correct", "description": "The slope angle matches the guide information and is <= 28 degrees.", "text": "**Slope Angle Correct:** The slope angle matches the guide information and is <= 28 degrees.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather information.", "steps": [{"name": "Wasatch Guide Visited", "description": "The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "text": "**Wasatch Guide Visited:** The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Utah Avalanche Center Visited", "description": "The browsing history contains a visit to the Utah Avalanche Center website.", "text": "**Utah Avalanche Center Visited:** The browsing history contains a visit to the Utah Avalanche Center website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Run Links Visited", "description": "At least one of the run links from the spreadsheet appears in the browsing history.", "text": "**Run Links Visited:** At least one of the run links from the spreadsheet appears in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Forecast Page Visited", "description": "The browsing history contains a visit to the avalanche forecast page for the relevant date.", "text": "**Forecast Page Visited:** The browsing history contains a visit to the avalanche forecast page for the relevant date.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Avalanche Forecast Data", "description": "The avalanche forecast information is correctly included and formatted.", "steps": [{"name": "Forecast Date Correct", "description": "The forecast date is 02/12/2025.", "text": "**Forecast Date Correct:** The forecast date is 02/12/2025.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Forecast Link Valid", "description": "The forecast link is a valid URL to the Utah Avalanche Center forecast.", "text": "**Forecast Link Valid:** The forecast link is a valid URL to the Utah Avalanche Center forecast.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Merged Cells", "description": "The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "text": "**Merged Cells:** The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Present", "description": "An image of the danger rose is present in the spreadsheet.", "text": "**Danger Rose Screenshot Present:** An image of the danger rose is present in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Image Valid", "description": "The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "text": "**Danger Rose Image Valid:** The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Danger Rating Cell Coloring", "description": "Each run name cell is colored according to its avalanche danger rating. (2 pts per run x 3 runs)", "steps": [{"name": "Danger Rating Determined", "description": "The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "text": "**Danger Rating Determined:** The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Cell Color Applied", "description": "The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "text": "**Cell Color Applied:** The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 32, "total_points": 50, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_25_skitourplan.4", "short_id": "25d", "family": "sheets_25_skitourplan", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Outdoor Activity Planner", "prompt": "Create a ski tour plan for 5 possible runs 25 degrees or less using the wasatch backcountry ski guide. Organize the route information for each run in a google sheet including the run name, run link, starting location, gps coordinates of the run location, typical vertical of the run, slope aspect, and slope angle as columns. Then look up the danger color rating for this run according to the Utah Avalanche Center forecast for 01/22/2025. Add columns for forecast date and forecast link and a screenshot of the full danger rose. These 3 columns should be merged vertically since they apply to all runs. Then using this information, determine the danger color rating for each run. Color the cell of each run name with the color of its danger rating.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns for ski run information and avalanche forecast data.", "steps": [{"name": "Run Name Column", "description": "There is a column for the run name.", "text": "**Run Name Column:** There is a column for the run name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Run Link Column", "description": "There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "text": "**Run Link Column:** There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Starting Location Column", "description": "There is a column for the starting location.", "text": "**Starting Location Column:** There is a column for the starting location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "GPS Coordinates Column", "description": "There is a column for GPS coordinates of the run location.", "text": "**GPS Coordinates Column:** There is a column for GPS coordinates of the run location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Typical Vertical Column", "description": "There is a column for typical vertical of the run.", "text": "**Typical Vertical Column:** There is a column for typical vertical of the run.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Aspect Column", "description": "There is a column for slope aspect.", "text": "**Slope Aspect Column:** There is a column for slope aspect.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Column", "description": "There is a column for slope angle.", "text": "**Slope Angle Column:** There is a column for slope angle.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Date Column", "description": "There is a column for forecast date (merged vertically).", "text": "**Forecast Date Column:** There is a column for forecast date (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Link Column", "description": "There is a column for forecast link (merged vertically).", "text": "**Forecast Link Column:** There is a column for forecast link (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Column", "description": "There is a column for the danger rose screenshot (merged vertically).", "text": "**Danger Rose Screenshot Column:** There is a column for the danger rose screenshot (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Run Selection Criteria", "description": "The spreadsheet contains exactly 5 ski runs that meet the slope angle requirement.", "steps": [{"name": "Run Count", "description": "The spreadsheet contains exactly 5 ski runs.", "text": "**Run Count:** The spreadsheet contains exactly 5 ski runs.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 1)", "description": "Run 1 has a slope angle of 25 degrees or less.", "text": "**Slope Angle Compliance (Run 1):** Run 1 has a slope angle of 25 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 2)", "description": "Run 2 has a slope angle of 25 degrees or less.", "text": "**Slope Angle Compliance (Run 2):** Run 2 has a slope angle of 25 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 3)", "description": "Run 3 has a slope angle of 25 degrees or less.", "text": "**Slope Angle Compliance (Run 3):** Run 3 has a slope angle of 25 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 4)", "description": "Run 4 has a slope angle of 25 degrees or less.", "text": "**Slope Angle Compliance (Run 4):** Run 4 has a slope angle of 25 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 5)", "description": "Run 5 has a slope angle of 25 degrees or less.", "text": "**Slope Angle Compliance (Run 5):** Run 5 has a slope angle of 25 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Run Data Accuracy", "description": "Each run's data matches information from the Wasatch Backcountry Ski Guide. (7 pts per run x 5 runs)", "steps": [{"name": "Run Name Valid", "description": "The run name exists in the Wasatch Backcountry Ski Guide.", "text": "**Run Name Valid:** The run name exists in the Wasatch Backcountry Ski Guide.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Run Link Valid", "description": "The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "text": "**Run Link Valid:** The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Starting Location Correct", "description": "The starting location matches the guide information.", "text": "**Starting Location Correct:** The starting location matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "GPS Coordinates Correct", "description": "The GPS coordinates match the guide information (within reasonable tolerance).", "text": "**GPS Coordinates Correct:** The GPS coordinates match the guide information (within reasonable tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Typical Vertical Correct", "description": "The typical vertical matches the guide information.", "text": "**Typical Vertical Correct:** The typical vertical matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Aspect Correct", "description": "The slope aspect matches the guide information.", "text": "**Slope Aspect Correct:** The slope aspect matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Correct", "description": "The slope angle matches the guide information and is <= 25 degrees.", "text": "**Slope Angle Correct:** The slope angle matches the guide information and is <= 25 degrees.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather information.", "steps": [{"name": "Wasatch Guide Visited", "description": "The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "text": "**Wasatch Guide Visited:** The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Utah Avalanche Center Visited", "description": "The browsing history contains a visit to the Utah Avalanche Center website.", "text": "**Utah Avalanche Center Visited:** The browsing history contains a visit to the Utah Avalanche Center website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Run Links Visited", "description": "At least one of the run links from the spreadsheet appears in the browsing history.", "text": "**Run Links Visited:** At least one of the run links from the spreadsheet appears in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Forecast Page Visited", "description": "The browsing history contains a visit to the avalanche forecast page for the relevant date.", "text": "**Forecast Page Visited:** The browsing history contains a visit to the avalanche forecast page for the relevant date.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Avalanche Forecast Data", "description": "The avalanche forecast information is correctly included and formatted.", "steps": [{"name": "Forecast Date Correct", "description": "The forecast date is 01/22/2025.", "text": "**Forecast Date Correct:** The forecast date is 01/22/2025.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Forecast Link Valid", "description": "The forecast link is a valid URL to the Utah Avalanche Center forecast.", "text": "**Forecast Link Valid:** The forecast link is a valid URL to the Utah Avalanche Center forecast.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Merged Cells", "description": "The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "text": "**Merged Cells:** The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Present", "description": "An image of the danger rose is present in the spreadsheet.", "text": "**Danger Rose Screenshot Present:** An image of the danger rose is present in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Image Valid", "description": "The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "text": "**Danger Rose Image Valid:** The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Danger Rating Cell Coloring", "description": "Each run name cell is colored according to its avalanche danger rating. (2 pts per run x 5 runs)", "steps": [{"name": "Danger Rating Determined", "description": "The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "text": "**Danger Rating Determined:** The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Cell Color Applied", "description": "The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "text": "**Cell Color Applied:** The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 34, "total_points": 70, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_25_skitourplan.5", "short_id": "25e", "family": "sheets_25_skitourplan", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Outdoor Activity Planner", "prompt": "Create a ski tour plan for 3 possible runs 32 degrees or less using the wasatch backcountry ski guide. Organize the route information for each run in a google sheet including the run name, run link, starting location, gps coordinates of the run location, typical vertical of the run, slope aspect, and slope angle as columns. Then look up the danger color rating for this run according to the Utah Avalanche Center forecast for 12/28/2024. Add columns for forecast date and forecast link and a screenshot of the full danger rose. These 3 columns should be merged vertically since they apply to all runs. Then using this information, determine the danger color rating for each run. Color the cell of each run name with the color of its danger rating.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns for ski run information and avalanche forecast data.", "steps": [{"name": "Run Name Column", "description": "There is a column for the run name.", "text": "**Run Name Column:** There is a column for the run name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Run Link Column", "description": "There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "text": "**Run Link Column:** There is a column for the run link (URL to Wasatch Backcountry Ski Guide).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Starting Location Column", "description": "There is a column for the starting location.", "text": "**Starting Location Column:** There is a column for the starting location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "GPS Coordinates Column", "description": "There is a column for GPS coordinates of the run location.", "text": "**GPS Coordinates Column:** There is a column for GPS coordinates of the run location.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Typical Vertical Column", "description": "There is a column for typical vertical of the run.", "text": "**Typical Vertical Column:** There is a column for typical vertical of the run.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Aspect Column", "description": "There is a column for slope aspect.", "text": "**Slope Aspect Column:** There is a column for slope aspect.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Column", "description": "There is a column for slope angle.", "text": "**Slope Angle Column:** There is a column for slope angle.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Date Column", "description": "There is a column for forecast date (merged vertically).", "text": "**Forecast Date Column:** There is a column for forecast date (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Forecast Link Column", "description": "There is a column for forecast link (merged vertically).", "text": "**Forecast Link Column:** There is a column for forecast link (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Column", "description": "There is a column for the danger rose screenshot (merged vertically).", "text": "**Danger Rose Screenshot Column:** There is a column for the danger rose screenshot (merged vertically).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Run Selection Criteria", "description": "The spreadsheet contains exactly 3 ski runs that meet the slope angle requirement.", "steps": [{"name": "Run Count", "description": "The spreadsheet contains exactly 3 ski runs.", "text": "**Run Count:** The spreadsheet contains exactly 3 ski runs.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 1)", "description": "Run 1 has a slope angle of 32 degrees or less.", "text": "**Slope Angle Compliance (Run 1):** Run 1 has a slope angle of 32 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 2)", "description": "Run 2 has a slope angle of 32 degrees or less.", "text": "**Slope Angle Compliance (Run 2):** Run 2 has a slope angle of 32 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Compliance (Run 3)", "description": "Run 3 has a slope angle of 32 degrees or less.", "text": "**Slope Angle Compliance (Run 3):** Run 3 has a slope angle of 32 degrees or less.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Run Data Accuracy", "description": "Each run's data matches information from the Wasatch Backcountry Ski Guide. (7 pts per run x 3 runs)", "steps": [{"name": "Run Name Valid", "description": "The run name exists in the Wasatch Backcountry Ski Guide.", "text": "**Run Name Valid:** The run name exists in the Wasatch Backcountry Ski Guide.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Run Link Valid", "description": "The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "text": "**Run Link Valid:** The run link is a valid URL to the Wasatch Backcountry Ski Guide page for this run.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Starting Location Correct", "description": "The starting location matches the guide information.", "text": "**Starting Location Correct:** The starting location matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "GPS Coordinates Correct", "description": "The GPS coordinates match the guide information (within reasonable tolerance).", "text": "**GPS Coordinates Correct:** The GPS coordinates match the guide information (within reasonable tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Typical Vertical Correct", "description": "The typical vertical matches the guide information.", "text": "**Typical Vertical Correct:** The typical vertical matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Aspect Correct", "description": "The slope aspect matches the guide information.", "text": "**Slope Aspect Correct:** The slope aspect matches the guide information.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Slope Angle Correct", "description": "The slope angle matches the guide information and is <= 32 degrees.", "text": "**Slope Angle Correct:** The slope angle matches the guide information and is <= 32 degrees.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather information.", "steps": [{"name": "Wasatch Guide Visited", "description": "The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "text": "**Wasatch Guide Visited:** The browsing history contains a visit to the Wasatch Backcountry Ski Guide website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Utah Avalanche Center Visited", "description": "The browsing history contains a visit to the Utah Avalanche Center website.", "text": "**Utah Avalanche Center Visited:** The browsing history contains a visit to the Utah Avalanche Center website.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Run Links Visited", "description": "At least one of the run links from the spreadsheet appears in the browsing history.", "text": "**Run Links Visited:** At least one of the run links from the spreadsheet appears in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": "Forecast Page Visited", "description": "The browsing history contains a visit to the avalanche forecast page for the relevant date.", "text": "**Forecast Page Visited:** The browsing history contains a visit to the avalanche forecast page for the relevant date.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Avalanche Forecast Data", "description": "The avalanche forecast information is correctly included and formatted.", "steps": [{"name": "Forecast Date Correct", "description": "The forecast date is 12/28/2024.", "text": "**Forecast Date Correct:** The forecast date is 12/28/2024.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Forecast Link Valid", "description": "The forecast link is a valid URL to the Utah Avalanche Center forecast.", "text": "**Forecast Link Valid:** The forecast link is a valid URL to the Utah Avalanche Center forecast.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": "Merged Cells", "description": "The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "text": "**Merged Cells:** The forecast date, forecast link, and danger rose screenshot columns are vertically merged across all run rows.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Danger Rose Screenshot Present", "description": "An image of the danger rose is present in the spreadsheet.", "text": "**Danger Rose Screenshot Present:** An image of the danger rose is present in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": "Danger Rose Image Valid", "description": "The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "text": "**Danger Rose Image Valid:** The danger rose screenshot shows the avalanche danger rose from the Utah Avalanche Center.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Danger Rating Cell Coloring", "description": "Each run name cell is colored according to its avalanche danger rating. (2 pts per run x 3 runs)", "steps": [{"name": "Danger Rating Determined", "description": "The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "text": "**Danger Rating Determined:** The danger color rating is correctly determined based on the run's slope aspect and angle relative to the danger rose.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": "Cell Color Applied", "description": "The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "text": "**Cell Color Applied:** The run name cell is colored with the appropriate danger rating color (green=Low, yellow=Moderate, orange=Considerable, red=High, black=Extreme).", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 32, "total_points": 50, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_25_skitourplan/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_28_personal_travel_planner.1", "short_id": "28a", "family": "sheets_28_personal_travel_planner", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Travel Planner", "prompt": "A family of 4 people is traveling to Washington D.C this May for 3 days, and they need a detailed day-by-day itinerary. Please prepare this travel plan and put everything into a Google Sheet so it’s easy to follow. They will stay at the same hotel Hyatt Regency Washington on Capitol Hill for all 3 days of the trip and use Public Transportation for travel during the trip. They plan to do a weekday trip.\n\nHere’s what they are looking for:\n- Activities & Sightseeing: Each day should have a mix of morning, afternoon, and evening plans at popular attractions in the city. Please include the opening hours, suggested duration, and any entrance fees. (For Activities & Sightseeing, there should be three rows per day: morning, afternoon, evening.)\n- Food Stops: Include two food stops per day (lunch and dinner). For each, add the type of cuisine, an average price estimate.\n- The food stops must be in the order: Lunch between Morning and Afternoon activity, Dinner between Afternoon and Evening activity.\n- Time of Day: Each row must specify the time slot — one of: Morning, Afternoon, Evening (for activities), or Lunch, Dinner (for food stops).\n- Start Time and Departure Time: Each row must include a start time (when the visit/meal begins) and a departure time (when the family leaves). These times must align with the Time of Day slot (e.g., Morning activities start in the morning hours, Lunch starts around midday). Start and departure times must not overlap with adjacent events and must leave enough time for transportation between destinations. The schedule across each day should flow sequentially without gaps or conflicts.\n- Transportation: For every entry (activity or food), note how they’d get between places (walking, metro, bus, taxi, etc.) and the typical travel time. Clearly indicate whether each day is a weekday or weekend, and adjust transit time estimates accordingly (e.g., account for rush hour delays on weekdays and reduced service or increased congestion on weekends). Ensure transit time correctness reflects these differences.\n- Costs: Provide an estimated cost for each main activity and meal, and color-code the cost column:\nunder $100 = green, $100–$200 = yellow, above $200 = orange\n- The review link for both Activity and Food should be from Google Maps.\n- Alternatives: For every destination (food or activity), suggest one backup option (for weather, crowds, or unexpected closures). Put these in the last column and format them in blue.\n- Please avoid repeating the same destination or restaurant across the trip.\n- If two places are far apart, adjust the order so the day makes geographic sense, and the pacing feels realistic (not jam-packed, but no big idle gaps).\n- Include the following columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost (with color coding), Alternative Option (in blue).\n- Freeze the top row (header).\n- Individual transit times should be <= 30 minutes and keep total transit times under 90 min per day", "checkpoints": [{"n": 1, "title": "Table Structure and Layout", "description": "The spreadsheet follows the required column structure, frozen header, and correct row-per-day breakdown.", "steps": [{"name": null, "description": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "text": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The header row is frozen, so it stays visible when scrolling.", "text": "The header row is frozen, so it stays visible when scrolling.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "text": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "text": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Content Completeness", "description": "Every activity and restaurant entry has all required fields filled in.", "steps": [{"name": null, "description": "Each activity/restaurant row has a non-empty destination name.", "text": "Each activity/restaurant row has a non-empty destination name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "text": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty opening hours value.", "text": "Each activity/restaurant row has a non-empty opening hours value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty start time.", "text": "Each activity/restaurant row has a non-empty start time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty departure time.", "text": "Each activity/restaurant row has a non-empty departure time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty duration value.", "text": "Each activity/restaurant row has a non-empty duration value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a valid Google Maps review link URL.", "text": "Each activity/restaurant row has a valid Google Maps review link URL.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty transportation mode.", "text": "Each activity/restaurant row has a non-empty transportation mode.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty travel time.", "text": "Each activity/restaurant row has a non-empty travel time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty cost value.", "text": "Each activity/restaurant row has a non-empty cost value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant row has a non-empty cuisine type.", "text": "Each restaurant row has a non-empty cuisine type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty alternative option.", "text": "Each activity/restaurant row has a non-empty alternative option.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty date column.", "text": "Each activity/restaurant row has a non-empty date column.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Destination Validity and Uniqueness", "description": "Every destination and alternative option refers to a real place in the target city, all are unique, and all attractions are open during their planned visit hours.", "steps": [{"name": null, "description": "Every activity and restaurant destination actually exists in the target city.", "text": "Every activity and restaurant destination actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every alternative option actually exists in the target city.", "text": "Every alternative option actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "text": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "All attractions are open during the planned visit hours.", "text": "All attractions are open during the planned visit hours.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Content Accuracy", "description": "All details are verified for factual accuracy against real-world data.", "steps": [{"name": null, "description": "Each destination's opening hours are verified as accurate.", "text": "Each destination's opening hours are verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each destination's entrance fee/cost is verified as accurate.", "text": "Each destination's entrance fee/cost is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every transportation mode is verified as available in the city for the given route.", "text": "Every transportation mode is verified as available in the city for the given route.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "text": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant entry's type of cuisine is verified as accurate.", "text": "Each restaurant entry's type of cuisine is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The review link URL contains actual reviews or information for the specific place mentioned.", "text": "The review link URL contains actual reviews or information for the specific place mentioned.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each alternative option is a viable substitute and is within a reachable distance.", "text": "Each alternative option is a viable substitute and is within a reachable distance.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "Conditional Formatting and Styling", "description": "Cost cells are color-coded by price range and alternative options are formatted in blue.", "steps": [{"name": null, "description": "The Cost column is correctly conditionally formatted:", "text": "The Cost column is correctly conditionally formatted:", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The Alternative Option column's text is formatted in blue.", "text": "The Alternative Option column's text is formatted in blue.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Logical Scheduling and Route Planning", "description": "The itinerary follows a sensible daily order with realistic timing, minimal backtracking, well-placed meal stops, feasible start/departure times, correct weekday/weekend transit adjustments, and transit time limits.", "steps": [{"name": null, "description": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "text": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "text": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "text": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "text": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "text": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "text": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "text": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 37, "total_points": 370, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_28_personal_travel_planner.2", "short_id": "28b", "family": "sheets_28_personal_travel_planner", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Travel Planner", "prompt": "A family of 5 people is traveling to Cairo this July for 5 days, and they need a detailed day-by-day itinerary. Please prepare this travel plan and put everything into a Google Sheet so it’s easy to follow. They will stay at the same hotel Marriott Mena House for all 5 days of the trip and use Public Transportation for travel during the trip. They plan to do a weekend trip.\n\nHere’s what they are looking for:\n- Activities & Sightseeing: Each day should have a mix of morning, afternoon, and evening plans at popular attractions in the city. Please include the opening hours, suggested duration, and any entrance fees. (For Activities & Sightseeing, there should be three rows per day: morning, afternoon, evening.)\n- Food Stops: Include two food stops per day (lunch and dinner). For each, add the type of cuisine, an average price estimate.\n- The food stops must be in the order: Lunch between Morning and Afternoon activity, Dinner between Afternoon and Evening activity.\n- Time of Day: Each row must specify the time slot — one of: Morning, Afternoon, Evening (for activities), or Lunch, Dinner (for food stops).\n- Start Time and Departure Time: Each row must include a start time (when the visit/meal begins) and a departure time (when the family leaves). These times must align with the Time of Day slot (e.g., Morning activities start in the morning hours, Lunch starts around midday). Start and departure times must not overlap with adjacent events and must leave enough time for transportation between destinations. The schedule across each day should flow sequentially without gaps or conflicts.\n- Transportation: For every entry (activity or food), note how they’d get between places (walking, metro, bus, taxi, etc.) and the typical travel time. Clearly indicate whether each day is a weekday or weekend, and adjust transit time estimates accordingly (e.g., account for rush hour delays on weekdays and reduced service or increased congestion on weekends). Ensure transit time correctness reflects these differences.\n- Costs: Provide an estimated cost for each main activity and meal, and color-code the cost column:\nunder $100 = green, $100–$200 = yellow, above $200 = orange\n- The review link for both Activity and Food should be from Google Maps.\n- Alternatives: For every destination (food or activity), suggest one backup option (for weather, crowds, or unexpected closures). Put these in the last column and format them in blue.\n- Please avoid repeating the same destination or restaurant across the trip.\n- If two places are far apart, adjust the order so the day makes geographic sense, and the pacing feels realistic (not jam-packed, but no big idle gaps).\n- Include the following columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost (with color coding), Alternative Option (in blue).\n- Freeze the top row (header).\n- Individual transit times should be <= 30 minutes and keep total transit times under 120 min per day", "checkpoints": [{"n": 1, "title": "Table Structure and Layout", "description": "The spreadsheet follows the required column structure, frozen header, and correct row-per-day breakdown.", "steps": [{"name": null, "description": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "text": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The header row is frozen, so it stays visible when scrolling.", "text": "The header row is frozen, so it stays visible when scrolling.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "text": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "text": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Content Completeness", "description": "Every activity and restaurant entry has all required fields filled in.", "steps": [{"name": null, "description": "Each activity/restaurant row has a non-empty destination name.", "text": "Each activity/restaurant row has a non-empty destination name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "text": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty opening hours value.", "text": "Each activity/restaurant row has a non-empty opening hours value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty start time.", "text": "Each activity/restaurant row has a non-empty start time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty departure time.", "text": "Each activity/restaurant row has a non-empty departure time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty duration value.", "text": "Each activity/restaurant row has a non-empty duration value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a valid Google Maps review link URL.", "text": "Each activity/restaurant row has a valid Google Maps review link URL.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty transportation mode.", "text": "Each activity/restaurant row has a non-empty transportation mode.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty travel time.", "text": "Each activity/restaurant row has a non-empty travel time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty cost value.", "text": "Each activity/restaurant row has a non-empty cost value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant row has a non-empty cuisine type.", "text": "Each restaurant row has a non-empty cuisine type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty alternative option.", "text": "Each activity/restaurant row has a non-empty alternative option.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty date column.", "text": "Each activity/restaurant row has a non-empty date column.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Destination Validity and Uniqueness", "description": "Every destination and alternative option refers to a real place in the target city, all are unique, and all attractions are open during their planned visit hours.", "steps": [{"name": null, "description": "Every activity and restaurant destination actually exists in the target city.", "text": "Every activity and restaurant destination actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every alternative option actually exists in the target city.", "text": "Every alternative option actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "text": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "All attractions are open during the planned visit hours.", "text": "All attractions are open during the planned visit hours.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Content Accuracy", "description": "All details are verified for factual accuracy against real-world data.", "steps": [{"name": null, "description": "Each destination's opening hours are verified as accurate.", "text": "Each destination's opening hours are verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each destination's entrance fee/cost is verified as accurate.", "text": "Each destination's entrance fee/cost is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every transportation mode is verified as available in the city for the given route.", "text": "Every transportation mode is verified as available in the city for the given route.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "text": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant entry's type of cuisine is verified as accurate.", "text": "Each restaurant entry's type of cuisine is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The review link URL contains actual reviews or information for the specific place mentioned.", "text": "The review link URL contains actual reviews or information for the specific place mentioned.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each alternative option is a viable substitute and is within a reachable distance.", "text": "Each alternative option is a viable substitute and is within a reachable distance.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "Conditional Formatting and Styling", "description": "Cost cells are color-coded by price range and alternative options are formatted in blue.", "steps": [{"name": null, "description": "The Cost column is correctly conditionally formatted:", "text": "The Cost column is correctly conditionally formatted:", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The Alternative Option column's text is formatted in blue.", "text": "The Alternative Option column's text is formatted in blue.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Logical Scheduling and Route Planning", "description": "The itinerary follows a sensible daily order with realistic timing, minimal backtracking, well-placed meal stops, feasible start/departure times, correct weekday/weekend transit adjustments, and transit time limits.", "steps": [{"name": null, "description": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "text": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "text": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "text": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "text": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "text": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "text": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Individual transit times are ≤ 30 minutes and total transit time per day is under 120 minutes.", "text": "Individual transit times are ≤ 30 minutes and total transit time per day is under 120 minutes.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 37, "total_points": 370, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_28_personal_travel_planner.3", "short_id": "28c", "family": "sheets_28_personal_travel_planner", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Travel Planner", "prompt": "A family of 8 people is traveling to Mexico City this August for 8 days, and they need a detailed day-by-day itinerary. Please prepare this travel plan and put everything into a Google Sheet so it’s easy to follow. They will stay at the same hotel Four Seasons Hotel Mexico City for all 8 days of the trip and use Public Transportation for travel during the trip. They plan to do a mixed trip.\n\nHere’s what they are looking for:\n- Activities & Sightseeing: Each day should have a mix of morning, afternoon, and evening plans at popular attractions in the city. Please include the opening hours, suggested duration, and any entrance fees. (For Activities & Sightseeing, there should be three rows per day: morning, afternoon, evening.)\n- Food Stops: Include two food stops per day (lunch and dinner). For each, add the type of cuisine, an average price estimate.\n- The food stops must be in the order: Lunch between Morning and Afternoon activity, Dinner between Afternoon and Evening activity.\n- Time of Day: Each row must specify the time slot — one of: Morning, Afternoon, Evening (for activities), or Lunch, Dinner (for food stops).\n- Start Time and Departure Time: Each row must include a start time (when the visit/meal begins) and a departure time (when the family leaves). These times must align with the Time of Day slot (e.g., Morning activities start in the morning hours, Lunch starts around midday). Start and departure times must not overlap with adjacent events and must leave enough time for transportation between destinations. The schedule across each day should flow sequentially without gaps or conflicts.\n- Transportation: For every entry (activity or food), note how they’d get between places (walking, metro, bus, taxi, etc.) and the typical travel time. Clearly indicate whether each day is a weekday or weekend, and adjust transit time estimates accordingly (e.g., account for rush hour delays on weekdays and reduced service or increased congestion on weekends). Ensure transit time correctness reflects these differences.\n- Costs: Provide an estimated cost for each main activity and meal, and color-code the cost column:\nunder $100 = green, $100–$200 = yellow, above $200 = orange\n- The review link for both Activity and Food should be from Google Maps.\n- Alternatives: For every destination (food or activity), suggest one backup option (for weather, crowds, or unexpected closures). Put these in the last column and format them in blue.\n- Please avoid repeating the same destination or restaurant across the trip.\n- If two places are far apart, adjust the order so the day makes geographic sense, and the pacing feels realistic (not jam-packed, but no big idle gaps).\n- Include the following columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost (with color coding), Alternative Option (in blue).\n- Freeze the top row (header).\n- Individual transit times should be <= 30 minutes and keep total transit times under 90 min per day", "checkpoints": [{"n": 1, "title": "Table Structure and Layout", "description": "The spreadsheet follows the required column structure, frozen header, and correct row-per-day breakdown.", "steps": [{"name": null, "description": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "text": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The header row is frozen, so it stays visible when scrolling.", "text": "The header row is frozen, so it stays visible when scrolling.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "text": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "text": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Content Completeness", "description": "Every activity and restaurant entry has all required fields filled in.", "steps": [{"name": null, "description": "Each activity/restaurant row has a non-empty destination name.", "text": "Each activity/restaurant row has a non-empty destination name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "text": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty opening hours value.", "text": "Each activity/restaurant row has a non-empty opening hours value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty start time.", "text": "Each activity/restaurant row has a non-empty start time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty departure time.", "text": "Each activity/restaurant row has a non-empty departure time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty duration value.", "text": "Each activity/restaurant row has a non-empty duration value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a valid Google Maps review link URL.", "text": "Each activity/restaurant row has a valid Google Maps review link URL.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty transportation mode.", "text": "Each activity/restaurant row has a non-empty transportation mode.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty travel time.", "text": "Each activity/restaurant row has a non-empty travel time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty cost value.", "text": "Each activity/restaurant row has a non-empty cost value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant row has a non-empty cuisine type.", "text": "Each restaurant row has a non-empty cuisine type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty alternative option.", "text": "Each activity/restaurant row has a non-empty alternative option.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty date column.", "text": "Each activity/restaurant row has a non-empty date column.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Destination Validity and Uniqueness", "description": "Every destination and alternative option refers to a real place in the target city, all are unique, and all attractions are open during their planned visit hours.", "steps": [{"name": null, "description": "Every activity and restaurant destination actually exists in the target city.", "text": "Every activity and restaurant destination actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every alternative option actually exists in the target city.", "text": "Every alternative option actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "text": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "All attractions are open during the planned visit hours.", "text": "All attractions are open during the planned visit hours.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Content Accuracy", "description": "All details are verified for factual accuracy against real-world data.", "steps": [{"name": null, "description": "Each destination's opening hours are verified as accurate.", "text": "Each destination's opening hours are verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each destination's entrance fee/cost is verified as accurate.", "text": "Each destination's entrance fee/cost is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every transportation mode is verified as available in the city for the given route.", "text": "Every transportation mode is verified as available in the city for the given route.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "text": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant entry's type of cuisine is verified as accurate.", "text": "Each restaurant entry's type of cuisine is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The review link URL contains actual reviews or information for the specific place mentioned.", "text": "The review link URL contains actual reviews or information for the specific place mentioned.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each alternative option is a viable substitute and is within a reachable distance.", "text": "Each alternative option is a viable substitute and is within a reachable distance.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "Conditional Formatting and Styling", "description": "Cost cells are color-coded by price range and alternative options are formatted in blue.", "steps": [{"name": null, "description": "The Cost column is correctly conditionally formatted:", "text": "The Cost column is correctly conditionally formatted:", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The Alternative Option column's text is formatted in blue.", "text": "The Alternative Option column's text is formatted in blue.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Logical Scheduling and Route Planning", "description": "The itinerary follows a sensible daily order with realistic timing, minimal backtracking, well-placed meal stops, feasible start/departure times, correct weekday/weekend transit adjustments, and transit time limits.", "steps": [{"name": null, "description": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "text": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "text": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "text": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "text": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "text": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "text": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "text": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 37, "total_points": 370, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_28_personal_travel_planner.4", "short_id": "28d", "family": "sheets_28_personal_travel_planner", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Travel Planner", "prompt": "A family of 6 people is traveling to Istanbul this December for 5 days, and they need a detailed day-by-day itinerary. Please prepare this travel plan and put everything into a Google Sheet so it’s easy to follow. They will stay at the same hotel Çırağan Palace Kempinski for all 5 days of the trip and use Public Transportation for travel during the trip. They plan to do a weekday trip.\n\nHere’s what they are looking for:\n- Activities & Sightseeing: Each day should have a mix of morning, afternoon, and evening plans at popular attractions in the city. Please include the opening hours, suggested duration, and any entrance fees. (For Activities & Sightseeing, there should be three rows per day: morning, afternoon, evening.)\n- Food Stops: Include two food stops per day (lunch and dinner). For each, add the type of cuisine, an average price estimate.\n- The food stops must be in the order: Lunch between Morning and Afternoon activity, Dinner between Afternoon and Evening activity.\n- Time of Day: Each row must specify the time slot — one of: Morning, Afternoon, Evening (for activities), or Lunch, Dinner (for food stops).\n- Start Time and Departure Time: Each row must include a start time (when the visit/meal begins) and a departure time (when the family leaves). These times must align with the Time of Day slot (e.g., Morning activities start in the morning hours, Lunch starts around midday). Start and departure times must not overlap with adjacent events and must leave enough time for transportation between destinations. The schedule across each day should flow sequentially without gaps or conflicts.\n- Transportation: For every entry (activity or food), note how they’d get between places (walking, metro, bus, taxi, etc.) and the typical travel time. Clearly indicate whether each day is a weekday or weekend, and adjust transit time estimates accordingly (e.g., account for rush hour delays on weekdays and reduced service or increased congestion on weekends). Ensure transit time correctness reflects these differences.\n- Costs: Provide an estimated cost for each main activity and meal, and color-code the cost column:\nunder $100 = green, $100–$200 = yellow, above $200 = orange\n- The review link for both Activity and Food should be from Google Maps.\n- Alternatives: For every destination (food or activity), suggest one backup option (for weather, crowds, or unexpected closures). Put these in the last column and format them in blue.\n- Please avoid repeating the same destination or restaurant across the trip.\n- If two places are far apart, adjust the order so the day makes geographic sense, and the pacing feels realistic (not jam-packed, but no big idle gaps).\n- Include the following columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost (with color coding), Alternative Option (in blue).\n- Freeze the top row (header).\n- Individual transit times should be <= 30 minutes and keep total transit times under 90 min per day", "checkpoints": [{"n": 1, "title": "Table Structure and Layout", "description": "The spreadsheet follows the required column structure, frozen header, and correct row-per-day breakdown.", "steps": [{"name": null, "description": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "text": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The header row is frozen, so it stays visible when scrolling.", "text": "The header row is frozen, so it stays visible when scrolling.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "text": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "text": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Content Completeness", "description": "Every activity and restaurant entry has all required fields filled in.", "steps": [{"name": null, "description": "Each activity/restaurant row has a non-empty destination name.", "text": "Each activity/restaurant row has a non-empty destination name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "text": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty opening hours value.", "text": "Each activity/restaurant row has a non-empty opening hours value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty start time.", "text": "Each activity/restaurant row has a non-empty start time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty departure time.", "text": "Each activity/restaurant row has a non-empty departure time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty duration value.", "text": "Each activity/restaurant row has a non-empty duration value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a valid Google Maps review link URL.", "text": "Each activity/restaurant row has a valid Google Maps review link URL.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty transportation mode.", "text": "Each activity/restaurant row has a non-empty transportation mode.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty travel time.", "text": "Each activity/restaurant row has a non-empty travel time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty cost value.", "text": "Each activity/restaurant row has a non-empty cost value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant row has a non-empty cuisine type.", "text": "Each restaurant row has a non-empty cuisine type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty alternative option.", "text": "Each activity/restaurant row has a non-empty alternative option.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty date column.", "text": "Each activity/restaurant row has a non-empty date column.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Destination Validity and Uniqueness", "description": "Every destination and alternative option refers to a real place in the target city, all are unique, and all attractions are open during their planned visit hours.", "steps": [{"name": null, "description": "Every activity and restaurant destination actually exists in the target city.", "text": "Every activity and restaurant destination actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every alternative option actually exists in the target city.", "text": "Every alternative option actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "text": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "All attractions are open during the planned visit hours.", "text": "All attractions are open during the planned visit hours.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Content Accuracy", "description": "All details are verified for factual accuracy against real-world data.", "steps": [{"name": null, "description": "Each destination's opening hours are verified as accurate.", "text": "Each destination's opening hours are verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each destination's entrance fee/cost is verified as accurate.", "text": "Each destination's entrance fee/cost is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every transportation mode is verified as available in the city for the given route.", "text": "Every transportation mode is verified as available in the city for the given route.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "text": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant entry's type of cuisine is verified as accurate.", "text": "Each restaurant entry's type of cuisine is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The review link URL contains actual reviews or information for the specific place mentioned.", "text": "The review link URL contains actual reviews or information for the specific place mentioned.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each alternative option is a viable substitute and is within a reachable distance.", "text": "Each alternative option is a viable substitute and is within a reachable distance.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "Conditional Formatting and Styling", "description": "Cost cells are color-coded by price range and alternative options are formatted in blue.", "steps": [{"name": null, "description": "The Cost column is correctly conditionally formatted:", "text": "The Cost column is correctly conditionally formatted:", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The Alternative Option column's text is formatted in blue.", "text": "The Alternative Option column's text is formatted in blue.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Logical Scheduling and Route Planning", "description": "The itinerary follows a sensible daily order with realistic timing, minimal backtracking, well-placed meal stops, feasible start/departure times, correct weekday/weekend transit adjustments, and transit time limits.", "steps": [{"name": null, "description": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "text": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "text": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "text": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "text": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "text": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "text": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "text": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 37, "total_points": 370, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_28_personal_travel_planner.5", "short_id": "28e", "family": "sheets_28_personal_travel_planner", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Travel Planner", "prompt": "A family of 7 people is traveling to Marrakech this March for 4 days, and they need a detailed day-by-day itinerary. Please prepare this travel plan and put everything into a Google Sheet so it’s easy to follow. They will stay at the same hotel La Mamounia for all 4 days of the trip and use Public Transportation for travel during the trip. They plan to do a weekend trip.\n\nHere’s what they are looking for:\n- Activities & Sightseeing: Each day should have a mix of morning, afternoon, and evening plans at popular attractions in the city. Please include the opening hours, suggested duration, and any entrance fees. (For Activities & Sightseeing, there should be three rows per day: morning, afternoon, evening.)\n- Food Stops: Include two food stops per day (lunch and dinner). For each, add the type of cuisine, an average price estimate.\n- The food stops must be in the order: Lunch between Morning and Afternoon activity, Dinner between Afternoon and Evening activity.\n- Time of Day: Each row must specify the time slot — one of: Morning, Afternoon, Evening (for activities), or Lunch, Dinner (for food stops).\n- Start Time and Departure Time: Each row must include a start time (when the visit/meal begins) and a departure time (when the family leaves). These times must align with the Time of Day slot (e.g., Morning activities start in the morning hours, Lunch starts around midday). Start and departure times must not overlap with adjacent events and must leave enough time for transportation between destinations. The schedule across each day should flow sequentially without gaps or conflicts.\n- Transportation: For every entry (activity or food), note how they’d get between places (walking, metro, bus, taxi, etc.) and the typical travel time. Clearly indicate whether each day is a weekday or weekend, and adjust transit time estimates accordingly (e.g., account for rush hour delays on weekdays and reduced service or increased congestion on weekends). Ensure transit time correctness reflects these differences.\n- Costs: Provide an estimated cost for each main activity and meal, and color-code the cost column:\nunder $100 = green, $100–$200 = yellow, above $200 = orange\n- The review link for both Activity and Food should be from Google Maps.\n- Alternatives: For every destination (food or activity), suggest one backup option (for weather, crowds, or unexpected closures). Put these in the last column and format them in blue.\n- Please avoid repeating the same destination or restaurant across the trip.\n- If two places are far apart, adjust the order so the day makes geographic sense, and the pacing feels realistic (not jam-packed, but no big idle gaps).\n- Include the following columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost (with color coding), Alternative Option (in blue).\n- Freeze the top row (header).\n- Individual transit times should be <= 30 minutes and keep total transit times under 90 min per day", "checkpoints": [{"n": 1, "title": "Table Structure and Layout", "description": "The spreadsheet follows the required column structure, frozen header, and correct row-per-day breakdown.", "steps": [{"name": null, "description": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "text": "The Google Sheet has all headers matching the requested columns: Date, Time of Day, Destination (Food/Activity), Cuisine, Opening time, Start Time, Departure Time, Duration, Review link, Transportation Mode, Travel Time, Cost, and Alternative Option (in blue).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The header row is frozen, so it stays visible when scrolling.", "text": "The header row is frozen, so it stays visible when scrolling.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "text": "The sheet contains exactly 3 rows per day for activities/sightseeing (morning, afternoon, evening).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "text": "The sheet contains exactly 2 rows per day for food stops (lunch and dinner).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Content Completeness", "description": "Every activity and restaurant entry has all required fields filled in.", "steps": [{"name": null, "description": "Each activity/restaurant row has a non-empty destination name.", "text": "Each activity/restaurant row has a non-empty destination name.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "text": "Each activity/restaurant row has a non-empty time of day value (e.g., Morning, Afternoon, Evening, Lunch, Dinner).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty opening hours value.", "text": "Each activity/restaurant row has a non-empty opening hours value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty start time.", "text": "Each activity/restaurant row has a non-empty start time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty departure time.", "text": "Each activity/restaurant row has a non-empty departure time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty duration value.", "text": "Each activity/restaurant row has a non-empty duration value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a valid Google Maps review link URL.", "text": "Each activity/restaurant row has a valid Google Maps review link URL.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty transportation mode.", "text": "Each activity/restaurant row has a non-empty transportation mode.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty travel time.", "text": "Each activity/restaurant row has a non-empty travel time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty cost value.", "text": "Each activity/restaurant row has a non-empty cost value.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant row has a non-empty cuisine type.", "text": "Each restaurant row has a non-empty cuisine type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty alternative option.", "text": "Each activity/restaurant row has a non-empty alternative option.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each activity/restaurant row has a non-empty date column.", "text": "Each activity/restaurant row has a non-empty date column.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Destination Validity and Uniqueness", "description": "Every destination and alternative option refers to a real place in the target city, all are unique, and all attractions are open during their planned visit hours.", "steps": [{"name": null, "description": "Every activity and restaurant destination actually exists in the target city.", "text": "Every activity and restaurant destination actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every alternative option actually exists in the target city.", "text": "Every alternative option actually exists in the target city.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "text": "All destinations and alternative options are unique — no destination or alternative option is repeated across the entire itinerary.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "All attractions are open during the planned visit hours.", "text": "All attractions are open during the planned visit hours.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Content Accuracy", "description": "All details are verified for factual accuracy against real-world data.", "steps": [{"name": null, "description": "Each destination's opening hours are verified as accurate.", "text": "Each destination's opening hours are verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each destination's entrance fee/cost is verified as accurate.", "text": "Each destination's entrance fee/cost is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every transportation mode is verified as available in the city for the given route.", "text": "Every transportation mode is verified as available in the city for the given route.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "text": "Every estimated travel time is verified as realistic for the given route and transportation mode.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each restaurant entry's type of cuisine is verified as accurate.", "text": "Each restaurant entry's type of cuisine is verified as accurate.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The review link URL contains actual reviews or information for the specific place mentioned.", "text": "The review link URL contains actual reviews or information for the specific place mentioned.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each alternative option is a viable substitute and is within a reachable distance.", "text": "Each alternative option is a viable substitute and is within a reachable distance.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "Conditional Formatting and Styling", "description": "Cost cells are color-coded by price range and alternative options are formatted in blue.", "steps": [{"name": null, "description": "The Cost column is correctly conditionally formatted:", "text": "The Cost column is correctly conditionally formatted:", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The Alternative Option column's text is formatted in blue.", "text": "The Alternative Option column's text is formatted in blue.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Logical Scheduling and Route Planning", "description": "The itinerary follows a sensible daily order with realistic timing, minimal backtracking, well-placed meal stops, feasible start/departure times, correct weekday/weekend transit adjustments, and transit time limits.", "steps": [{"name": null, "description": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "text": "Activities are arranged in a logical order, minimizing travel time and avoiding backtracking across the city.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "text": "Lunch is placed between the Morning and Afternoon activity, and Dinner is placed between the Afternoon and Evening activity each day.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "text": "Each row's duration is consistent with typical visit or dining times (e.g., 1–2 hours for lunch, 2–3 hours for museums).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "text": "Each row's start time is feasible: it falls within the correct time block for its Time of Day slot (e.g., Morning starts between 8–11 AM, Lunch between 11 AM–2 PM, Afternoon between 12–5 PM, Dinner between 5–9 PM, Evening between 5–10 PM), and does not start before the previous event's departure time plus its travel time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "text": "Each row's departure time is feasible: it comes after the start time, the gap between start and departure is consistent with the stated duration, and it does not overlap with the next event's start time.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "text": "Transit time estimates are adjusted accordingly for each day depending on whether it is a weekday or weekend (e.g., rush hour delays on weekdays, reduced service or increased congestion on weekends).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "text": "Individual transit times are ≤ 30 minutes and total transit time per day is under 90 minutes.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 37, "total_points": 370, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_28_personal_travel_planner/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_2_personal_recipe.1", "short_id": "2a", "family": "sheets_2_personal_recipe_foodcomposition", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Recipes", "prompt": "From the recipe https://rainbowplantlife.com/guide-to-making-the-best-cashew-cream/#ingredients, pull the average amount of iron, carbohydrates, fat, sugar, potassium, vitamin A, protein, sodium, fiber, calcium, and vitamin C for each ingredient (based on its quantity) using the USDA food database. When searching the USDA, use the food code that best matches the ingredient; more general codes are usually preferable, unless the recipe specifies a matching brand. For instance, if a listed ingredient is just “Paneer”, you would use “Cheese, paneer” instead of “Palak Paneer”. You may use any of the five USDA food search tabs. Place the results into a Google Sheet as follows. Group nutrients into macros, minerals, and vitamins. Within each group, sort nutrients alphabetically. The first column should be titled \"Ingredients\" and the second “Link”, followed by sorted nutrient columns starting with macros, then minerals, then vitamins. Above each group, add a merged, centered, italicized header cell labeled \"Macros\", \"Minerals\", or \"Vitamins\" spanning all associated columns. Bold all column titles, excluding the group header cells. Color all cells in each group with a light background color, making sure each group has a distinct color. Place a horizontal line under column titles. List ingredients in the first column in the order they appear in the recipe. In the “Link” column, add the USDA data link you used to obtain the quantity of a nutrient in a specific food. Bold any nutrient values exceeding 10% DV according to FDA guidelines. For ingredients you can’t find in the USDA database, leave centered N/A. For ingredients with a range of values as the amount, use the minimum value. Ignore ingredients that are listed as optional or don’t have a specified amount.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure & Column Layout", "description": "The spreadsheet contains the required columns in the correct order with proper nutrient groupings.", "steps": [{"name": null, "description": "\"Ingredients\" column exists and is first.", "text": "\"Ingredients\" column exists and is first.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Link\" column exists and is second.", "text": "\"Link\" column exists and is second.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Carbohydrates column present.", "text": "Carbohydrates column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fat column present.", "text": "Fat column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fiber column present.", "text": "Fiber column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Protein column present.", "text": "Protein column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sugar column present.", "text": "Sugar column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Calcium column present.", "text": "Calcium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Iron column present.", "text": "Iron column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Potassium column present.", "text": "Potassium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sodium column present.", "text": "Sodium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A column present.", "text": "Vitamin A column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C column present.", "text": "Vitamin C column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Macros columns are in alphabetical order.", "text": "Macros columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Minerals columns are in alphabetical order.", "text": "Minerals columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamins columns are in alphabetical order.", "text": "Vitamins columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Group Headers & Formatting", "description": "The spreadsheet has properly formatted group headers for nutrient categories.", "steps": [{"name": null, "description": "\"Macros\" merged header exists spanning macro columns.", "text": "\"Macros\" merged header exists spanning macro columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Minerals\" merged header exists spanning mineral columns.", "text": "\"Minerals\" merged header exists spanning mineral columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Vitamins\" merged header exists spanning vitamin columns.", "text": "\"Vitamins\" merged header exists spanning vitamin columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Group headers are centered and italicized.", "text": "Group headers are centered and italicized.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Column titles are bolded (excluding group headers).", "text": "Column titles are bolded (excluding group headers).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Horizontal line under column titles exists.", "text": "Horizontal line under column titles exists.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Color Formatting", "description": "Each nutrient group has a distinct background color applied to all cells in that group.", "steps": [{"name": null, "description": "Macro columns share same background color.", "text": "Macro columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Mineral columns share same background color.", "text": "Mineral columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin columns share same background color.", "text": "Vitamin columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Three groups have distinct colors from each other.", "text": "Three groups have distinct colors from each other.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Ingredients Present", "description": "The correct ingredients from the recipe are present in the spreadsheet, excluding those without specified amounts.", "steps": [{"name": null, "description": "Raw Cashews row present.", "text": "Raw Cashews row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Water row present.", "text": "Water row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Garlic row present.", "text": "Garlic row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sea Salt row present.", "text": "Sea Salt row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Nutritional Yeast row present.", "text": "Nutritional Yeast row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Lemon Juice row present.", "text": "Lemon Juice row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Onion Powder row present.", "text": "Onion Powder row present.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Black Pepper row NOT present (excluded - no specified amount).", "text": "Black Pepper row NOT present (excluded - no specified amount).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 5, "title": "USDA Links Validation", "description": "Each ingredient has a valid USDA FoodData Central link that corresponds to the correct food item.", "steps": [{"name": null, "description": "Raw Cashews link valid and matches ingredient (via HTML parsing).", "text": "Raw Cashews link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Water link valid and matches ingredient (via HTML parsing).", "text": "Water link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Garlic link valid and matches ingredient (via HTML parsing).", "text": "Garlic link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sea Salt link valid and matches ingredient (via HTML parsing).", "text": "Sea Salt link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Nutritional Yeast link valid and matches ingredient (via HTML parsing).", "text": "Nutritional Yeast link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Lemon Juice link valid and matches ingredient (via HTML parsing).", "text": "Lemon Juice link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Onion Powder link valid and matches ingredient (via HTML parsing).", "text": "Onion Powder link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Nutrient Values Accuracy", "description": "Nutrient values for each ingredient match the expected gold label values within tolerance. Each nutrient column is worth 7 points (one per ingredient).", "steps": [{"name": null, "description": "Carbohydrates values match gold labels within tolerance (7 pts).", "text": "Carbohydrates values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fat values match gold labels within tolerance (7 pts).", "text": "Fat values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fiber values match gold labels within tolerance (7 pts).", "text": "Fiber values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Protein values match gold labels within tolerance (7 pts).", "text": "Protein values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sugar values match gold labels within tolerance (7 pts).", "text": "Sugar values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Calcium values match gold labels within tolerance (7 pts).", "text": "Calcium values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Iron values match gold labels within tolerance (7 pts).", "text": "Iron values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Potassium values match gold labels within tolerance (7 pts).", "text": "Potassium values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sodium values match gold labels within tolerance (7 pts).", "text": "Sodium values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A values match gold labels within tolerance (7 pts).", "text": "Vitamin A values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C values match gold labels within tolerance (7 pts).", "text": "Vitamin C values match gold labels within tolerance (7 pts).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Bold Formatting for >10% DV", "description": "Nutrient values exceeding 10% of the FDA daily value are bolded.", "steps": [{"name": null, "description": "Values exceeding 10% DV are bolded per FDA guidelines.", "text": "Values exceeding 10% DV are bolded per FDA guidelines.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 8, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather recipe and nutritional information.", "steps": [{"name": null, "description": "Recipe URL (rainbowplantlife.com) visited (1 pt).", "text": "Recipe URL (rainbowplantlife.com) visited (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Raw Cashews (1 pt).", "text": "USDA database URL visited for Raw Cashews (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Water (1 pt).", "text": "USDA database URL visited for Water (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Garlic (1 pt).", "text": "USDA database URL visited for Garlic (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Sea Salt (1 pt).", "text": "USDA database URL visited for Sea Salt (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Nutritional Yeast (1 pt).", "text": "USDA database URL visited for Nutritional Yeast (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Lemon Juice (1 pt).", "text": "USDA database URL visited for Lemon Juice (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Onion Powder (1 pt).", "text": "USDA database URL visited for Onion Powder (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 8, "n_steps": 61, "total_points": 127, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_2_personal_recipe.2", "short_id": "2b", "family": "sheets_2_personal_recipe_foodcomposition", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Recipes", "prompt": "From the recipe https://bakingmischief.com/homemade-mocha/, pull the average amount of iron, carbohydrates, fat, sugar, potassium, vitamin A, protein, sodium, fiber, calcium, and vitamin C for each ingredient (based on its quantity) using the USDA food database. When searching the USDA, use the food code that best matches the ingredient; more general codes are usually preferable, unless the recipe specifies a matching brand. For instance, if a listed ingredient is just \"Paneer\", you would use \"Cheese, paneer\" instead of \"Palak Paneer\". You may use any of the five USDA food search tabs. Place the results into a Google Sheet as follows. Group nutrients into macros, minerals, and vitamins. Within each group, sort nutrients alphabetically. The first column should be titled \"Ingredients\" and the second \"Link\", followed by sorted nutrient columns starting with macros, then minerals, then vitamins. Above each group, add a merged, centered, italicized header cell labeled \"Macros\", \"Minerals\", or \"Vitamins\" spanning all associated columns. Bold all column titles, excluding the group header cells. Color all cells in each group with a light background color, making sure each group has a distinct color. Place a horizontal line under column titles. List ingredients in the first column in the order they appear in the recipe. In the \"Link\" column, add the USDA data link you used to obtain the quantity of a nutrient in a specific food. Bold any nutrient values exceeding 10% DV according to FDA guidelines. For ingredients you can't find in the USDA database, leave centered N/A. For ingredients with a range of values as the amount, use the minimum value. Ignore ingredients that are listed as optional or don't have a specified amount.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure & Column Layout", "description": "The spreadsheet contains the required columns in the correct order with proper nutrient groupings.", "steps": [{"name": null, "description": "\"Ingredients\" column exists and is first.", "text": "\"Ingredients\" column exists and is first.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Link\" column exists and is second.", "text": "\"Link\" column exists and is second.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Carbohydrates column present.", "text": "Carbohydrates column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fat column present.", "text": "Fat column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fiber column present.", "text": "Fiber column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Protein column present.", "text": "Protein column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sugar column present.", "text": "Sugar column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Calcium column present.", "text": "Calcium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Iron column present.", "text": "Iron column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Potassium column present.", "text": "Potassium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sodium column present.", "text": "Sodium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A column present.", "text": "Vitamin A column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C column present.", "text": "Vitamin C column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Macros columns are in alphabetical order.", "text": "Macros columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Minerals columns are in alphabetical order.", "text": "Minerals columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamins columns are in alphabetical order.", "text": "Vitamins columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Group Headers & Formatting", "description": "The spreadsheet has properly formatted group headers for nutrient categories.", "steps": [{"name": null, "description": "\"Macros\" merged header exists spanning macro columns.", "text": "\"Macros\" merged header exists spanning macro columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Minerals\" merged header exists spanning mineral columns.", "text": "\"Minerals\" merged header exists spanning mineral columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Vitamins\" merged header exists spanning vitamin columns.", "text": "\"Vitamins\" merged header exists spanning vitamin columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Group headers are centered and italicized.", "text": "Group headers are centered and italicized.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Column titles are bolded (excluding group headers).", "text": "Column titles are bolded (excluding group headers).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Horizontal line under column titles exists.", "text": "Horizontal line under column titles exists.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Color Formatting", "description": "Each nutrient group has a distinct background color applied to all cells in that group.", "steps": [{"name": null, "description": "Macro columns share same background color.", "text": "Macro columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Mineral columns share same background color.", "text": "Mineral columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin columns share same background color.", "text": "Vitamin columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Three groups have distinct colors from each other.", "text": "Three groups have distinct colors from each other.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Ingredients Present", "description": "The correct ingredients from the recipe are present in the spreadsheet, excluding those without specified amounts.", "steps": [{"name": null, "description": "granulated sugar row present.", "text": "granulated sugar row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "unsweetened cocoa powder row present.", "text": "unsweetened cocoa powder row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "hot coffee row present.", "text": "hot coffee row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "milk row present.", "text": "milk row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Water row present.", "text": "Water row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "whipped cream row NOT present (excluded - optional).", "text": "whipped cream row NOT present (excluded - optional).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "USDA Links Validation", "description": "Each ingredient has a valid USDA FoodData Central link that corresponds to the correct food item.", "steps": [{"name": null, "description": "granulated sugar link valid and matches ingredient (via HTML parsing).", "text": "granulated sugar link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "unsweetened cocoa powder link valid and matches ingredient (via HTML parsing).", "text": "unsweetened cocoa powder link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "hot coffee link valid and matches ingredient (via HTML parsing).", "text": "hot coffee link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "milk link valid and matches ingredient (via HTML parsing).", "text": "milk link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Water link valid and matches ingredient (via HTML parsing).", "text": "Water link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Nutrient Values Accuracy", "description": "Nutrient values for each ingredient match the expected gold label values within tolerance. Each nutrient column is worth 5 points (one per ingredient).", "steps": [{"name": null, "description": "Carbohydrates values match gold labels within tolerance (5 pts).", "text": "Carbohydrates values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fat values match gold labels within tolerance (5 pts).", "text": "Fat values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fiber values match gold labels within tolerance (5 pts).", "text": "Fiber values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Protein values match gold labels within tolerance (5 pts).", "text": "Protein values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sugar values match gold labels within tolerance (5 pts).", "text": "Sugar values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Calcium values match gold labels within tolerance (5 pts).", "text": "Calcium values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Iron values match gold labels within tolerance (5 pts).", "text": "Iron values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Potassium values match gold labels within tolerance (5 pts).", "text": "Potassium values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sodium values match gold labels within tolerance (5 pts).", "text": "Sodium values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A values match gold labels within tolerance (5 pts).", "text": "Vitamin A values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C values match gold labels within tolerance (5 pts).", "text": "Vitamin C values match gold labels within tolerance (5 pts).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Bold Formatting for >10% DV", "description": "Nutrient values exceeding 10% of the FDA daily value are bolded.", "steps": [{"name": null, "description": "Values exceeding 10% DV are bolded per FDA guidelines.", "text": "Values exceeding 10% DV are bolded per FDA guidelines.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 8, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather recipe and nutritional information.", "steps": [{"name": null, "description": "Recipe URL (bakingmischief.com) visited (1 pt).", "text": "Recipe URL (bakingmischief.com) visited (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for granulated sugar (1 pt).", "text": "USDA database URL visited for granulated sugar (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for unsweetened cocoa powder (1 pt).", "text": "USDA database URL visited for unsweetened cocoa powder (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for hot coffee (1 pt).", "text": "USDA database URL visited for hot coffee (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for milk (1 pt).", "text": "USDA database URL visited for milk (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Water (1 pt).", "text": "USDA database URL visited for Water (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 8, "n_steps": 55, "total_points": 99, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_2_personal_recipe.3", "short_id": "2c", "family": "sheets_2_personal_recipe_foodcomposition", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Recipes", "prompt": "From the recipe https://www.loveandlemons.com/guacamole-recipe/, pull the average amount of iron, carbohydrates, fat, sugar, potassium, vitamin A, protein, sodium, fiber, calcium, and vitamin C for each ingredient (based on its quantity) using the USDA food database. When searching the USDA, use the food code that best matches the ingredient; more general codes are usually preferable, unless the recipe specifies a matching brand. For instance, if a listed ingredient is just \"Paneer\", you would use \"Cheese, paneer\" instead of \"Palak Paneer\". You may use any of the five USDA food search tabs. Place the results into a Google Sheet as follows. Group nutrients into macros, minerals, and vitamins. Within each group, sort nutrients alphabetically. The first column should be titled \"Ingredients\" and the second \"Link\", followed by sorted nutrient columns starting with macros, then minerals, then vitamins. Above each group, add a merged, centered, italicized header cell labeled \"Macros\", \"Minerals\", or \"Vitamins\" spanning all associated columns. Bold all column titles, excluding the group header cells. Color all cells in each group with a light background color, making sure each group has a distinct color. Place a horizontal line under column titles. List ingredients in the first column in the order they appear in the recipe. In the \"Link\" column, add the USDA data link you used to obtain the quantity of a nutrient in a specific food. Bold any nutrient values exceeding 10% DV according to FDA guidelines. For ingredients you can't find in the USDA database, leave centered N/A. For ingredients with a range of values as the amount, use the minimum value. Ignore ingredients that are listed as optional or don't have a specified amount.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure & Column Layout", "description": "The spreadsheet contains the required columns in the correct order with proper nutrient groupings.", "steps": [{"name": null, "description": "\"Ingredients\" column exists and is first.", "text": "\"Ingredients\" column exists and is first.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Link\" column exists and is second.", "text": "\"Link\" column exists and is second.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Carbohydrates column present.", "text": "Carbohydrates column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fat column present.", "text": "Fat column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fiber column present.", "text": "Fiber column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Protein column present.", "text": "Protein column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sugar column present.", "text": "Sugar column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Calcium column present.", "text": "Calcium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Iron column present.", "text": "Iron column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Potassium column present.", "text": "Potassium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sodium column present.", "text": "Sodium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A column present.", "text": "Vitamin A column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C column present.", "text": "Vitamin C column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Macros columns are in alphabetical order.", "text": "Macros columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Minerals columns are in alphabetical order.", "text": "Minerals columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamins columns are in alphabetical order.", "text": "Vitamins columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Group Headers & Formatting", "description": "The spreadsheet has properly formatted group headers for nutrient categories.", "steps": [{"name": null, "description": "\"Macros\" merged header exists spanning macro columns.", "text": "\"Macros\" merged header exists spanning macro columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Minerals\" merged header exists spanning mineral columns.", "text": "\"Minerals\" merged header exists spanning mineral columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Vitamins\" merged header exists spanning vitamin columns.", "text": "\"Vitamins\" merged header exists spanning vitamin columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Group headers are centered and italicized.", "text": "Group headers are centered and italicized.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Column titles are bolded (excluding group headers).", "text": "Column titles are bolded (excluding group headers).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Horizontal line under column titles exists.", "text": "Horizontal line under column titles exists.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Color Formatting", "description": "Each nutrient group has a distinct background color applied to all cells in that group.", "steps": [{"name": null, "description": "Macro columns share same background color.", "text": "Macro columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Mineral columns share same background color.", "text": "Mineral columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin columns share same background color.", "text": "Vitamin columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Three groups have distinct colors from each other.", "text": "Three groups have distinct colors from each other.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Ingredients Present", "description": "The correct ingredients from the recipe are present in the spreadsheet, excluding those marked as optional.", "steps": [{"name": null, "description": "Avocados row present.", "text": "Avocados row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "red onion row present.", "text": "red onion row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "cilantro row present.", "text": "cilantro row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "lime juice row present.", "text": "lime juice row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "jalapeno pepper row present.", "text": "jalapeno pepper row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "sea salt row present.", "text": "sea salt row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "ground cumin row NOT present (excluded - optional).", "text": "ground cumin row NOT present (excluded - optional).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Roma tomatoes row NOT present (excluded - optional).", "text": "Roma tomatoes row NOT present (excluded - optional).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "USDA Links Validation", "description": "Each ingredient has a valid USDA FoodData Central link that corresponds to the correct food item.", "steps": [{"name": null, "description": "Avocados link valid and matches ingredient (via HTML parsing).", "text": "Avocados link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "red onion link valid and matches ingredient (via HTML parsing).", "text": "red onion link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "cilantro link valid and matches ingredient (via HTML parsing).", "text": "cilantro link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "lime juice link valid and matches ingredient (via HTML parsing).", "text": "lime juice link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "jalapeno pepper link valid and matches ingredient (via HTML parsing).", "text": "jalapeno pepper link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "sea salt link valid and matches ingredient (via HTML parsing).", "text": "sea salt link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Nutrient Values Accuracy", "description": "Nutrient values for each ingredient match the expected gold label values within tolerance. Each nutrient column is worth 6 points (one per ingredient).", "steps": [{"name": null, "description": "Carbohydrates values match gold labels within tolerance (6 pts).", "text": "Carbohydrates values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fat values match gold labels within tolerance (6 pts).", "text": "Fat values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fiber values match gold labels within tolerance (6 pts).", "text": "Fiber values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Protein values match gold labels within tolerance (6 pts).", "text": "Protein values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sugar values match gold labels within tolerance (6 pts).", "text": "Sugar values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Calcium values match gold labels within tolerance (6 pts).", "text": "Calcium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Iron values match gold labels within tolerance (6 pts).", "text": "Iron values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Potassium values match gold labels within tolerance (6 pts).", "text": "Potassium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sodium values match gold labels within tolerance (6 pts).", "text": "Sodium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A values match gold labels within tolerance (6 pts).", "text": "Vitamin A values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C values match gold labels within tolerance (6 pts).", "text": "Vitamin C values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Bold Formatting for >10% DV", "description": "Nutrient values exceeding 10% of the FDA daily value are bolded.", "steps": [{"name": null, "description": "Values exceeding 10% DV are bolded per FDA guidelines.", "text": "Values exceeding 10% DV are bolded per FDA guidelines.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 8, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather recipe and nutritional information.", "steps": [{"name": null, "description": "Recipe URL (loveandlemons.com) visited (1 pt).", "text": "Recipe URL (loveandlemons.com) visited (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for Avocados (1 pt).", "text": "USDA database URL visited for Avocados (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for red onion (1 pt).", "text": "USDA database URL visited for red onion (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for cilantro (1 pt).", "text": "USDA database URL visited for cilantro (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for lime juice (1 pt).", "text": "USDA database URL visited for lime juice (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for jalapeno pepper (1 pt).", "text": "USDA database URL visited for jalapeno pepper (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for sea salt (1 pt).", "text": "USDA database URL visited for sea salt (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 8, "n_steps": 59, "total_points": 114, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_2_personal_recipe.4", "short_id": "2d", "family": "sheets_2_personal_recipe_foodcomposition", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Recipes", "prompt": "From the recipe https://minimalistbaker.com/easy-vegan-fried-rice/, pull the average amount of iron, carbohydrates, fat, sugar, potassium, vitamin A, protein, sodium, fiber, calcium, and vitamin C for each ingredient (based on its quantity) using the USDA food database. When searching the USDA, use the food code that best matches the ingredient; more general codes are usually preferable, unless the recipe specifies a matching brand. For instance, if a listed ingredient is just \"Paneer\", you would use \"Cheese, paneer\" instead of \"Palak Paneer\". You may use any of the five USDA food search tabs. Place the results into a Google Sheet as follows. Group nutrients into macros, minerals, and vitamins. Within each group, sort nutrients alphabetically. The first column should be titled \"Ingredients\" and the second \"Link\", followed by sorted nutrient columns starting with macros, then minerals, then vitamins. Above each group, add a merged, centered, italicized header cell labeled \"Macros\", \"Minerals\", or \"Vitamins\" spanning all associated columns. Bold all column titles, excluding the group header cells. Color all cells in each group with a light background color, making sure each group has a distinct color. Place a horizontal line under column titles. List ingredients in the first column in the order they appear in the recipe. In the \"Link\" column, add the USDA data link you used to obtain the quantity of a nutrient in a specific food. Bold any nutrient values exceeding 10% DV according to FDA guidelines. For ingredients you can't find in the USDA database, leave centered N/A. For ingredients with a range of values as the amount, use the minimum value. Ignore ingredients that are listed as optional or don't have a specified amount.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure & Column Layout", "description": "The spreadsheet contains the required columns in the correct order with proper nutrient groupings.", "steps": [{"name": null, "description": "\"Ingredients\" column exists and is first.", "text": "\"Ingredients\" column exists and is first.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Link\" column exists and is second.", "text": "\"Link\" column exists and is second.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Carbohydrates column present.", "text": "Carbohydrates column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fat column present.", "text": "Fat column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fiber column present.", "text": "Fiber column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Protein column present.", "text": "Protein column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sugar column present.", "text": "Sugar column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Calcium column present.", "text": "Calcium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Iron column present.", "text": "Iron column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Potassium column present.", "text": "Potassium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sodium column present.", "text": "Sodium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A column present.", "text": "Vitamin A column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C column present.", "text": "Vitamin C column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Macros columns are in alphabetical order.", "text": "Macros columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Minerals columns are in alphabetical order.", "text": "Minerals columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamins columns are in alphabetical order.", "text": "Vitamins columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Group Headers & Formatting", "description": "The spreadsheet has properly formatted group headers for nutrient categories.", "steps": [{"name": null, "description": "\"Macros\" merged header exists spanning macro columns.", "text": "\"Macros\" merged header exists spanning macro columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Minerals\" merged header exists spanning mineral columns.", "text": "\"Minerals\" merged header exists spanning mineral columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Vitamins\" merged header exists spanning vitamin columns.", "text": "\"Vitamins\" merged header exists spanning vitamin columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Group headers are centered and italicized.", "text": "Group headers are centered and italicized.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Column titles are bolded (excluding group headers).", "text": "Column titles are bolded (excluding group headers).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Horizontal line under column titles exists.", "text": "Horizontal line under column titles exists.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Color Formatting", "description": "Each nutrient group has a distinct background color applied to all cells in that group.", "steps": [{"name": null, "description": "Macro columns share same background color.", "text": "Macro columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Mineral columns share same background color.", "text": "Mineral columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin columns share same background color.", "text": "Vitamin columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Three groups have distinct colors from each other.", "text": "Three groups have distinct colors from each other.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Ingredients Present", "description": "The correct ingredients from the recipe are present in the spreadsheet, excluding those marked as optional.", "steps": [{"name": null, "description": "extra-firm tofu row present.", "text": "extra-firm tofu row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "brown rice row present.", "text": "brown rice row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "garlic row present.", "text": "garlic row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "green onion row present.", "text": "green onion row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "peas row present.", "text": "peas row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "carrots row present.", "text": "carrots row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "soy sauce row present.", "text": "soy sauce row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "peanut butter row present.", "text": "peanut butter row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "brown sugar row present.", "text": "brown sugar row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "chili garlic sauce row present.", "text": "chili garlic sauce row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "toasted sesame oil row NOT present (excluded - optional).", "text": "toasted sesame oil row NOT present (excluded - optional).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "USDA Links Validation", "description": "Each ingredient has a valid USDA FoodData Central link that corresponds to the correct food item.", "steps": [{"name": null, "description": "extra-firm tofu link valid and matches ingredient (via HTML parsing).", "text": "extra-firm tofu link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "brown rice link valid and matches ingredient (via HTML parsing).", "text": "brown rice link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "garlic link valid and matches ingredient (via HTML parsing).", "text": "garlic link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "green onion link valid and matches ingredient (via HTML parsing).", "text": "green onion link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "peas link valid and matches ingredient (via HTML parsing).", "text": "peas link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "carrots link valid and matches ingredient (via HTML parsing).", "text": "carrots link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "soy sauce link valid and matches ingredient (via HTML parsing).", "text": "soy sauce link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "peanut butter link valid and matches ingredient (via HTML parsing).", "text": "peanut butter link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "brown sugar link valid and matches ingredient (via HTML parsing).", "text": "brown sugar link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "chili garlic sauce link valid and matches ingredient (via HTML parsing).", "text": "chili garlic sauce link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Nutrient Values Accuracy", "description": "Nutrient values for each ingredient match the expected gold label values within tolerance. Each nutrient column is worth 10 points (one per ingredient).", "steps": [{"name": null, "description": "Carbohydrates values match gold labels within tolerance (10 pts).", "text": "Carbohydrates values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fat values match gold labels within tolerance (10 pts).", "text": "Fat values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fiber values match gold labels within tolerance (10 pts).", "text": "Fiber values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Protein values match gold labels within tolerance (10 pts).", "text": "Protein values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sugar values match gold labels within tolerance (10 pts).", "text": "Sugar values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Calcium values match gold labels within tolerance (10 pts).", "text": "Calcium values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Iron values match gold labels within tolerance (10 pts).", "text": "Iron values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Potassium values match gold labels within tolerance (10 pts).", "text": "Potassium values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sodium values match gold labels within tolerance (10 pts).", "text": "Sodium values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A values match gold labels within tolerance (10 pts).", "text": "Vitamin A values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C values match gold labels within tolerance (10 pts).", "text": "Vitamin C values match gold labels within tolerance (10 pts).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Bold Formatting for >10% DV", "description": "Nutrient values exceeding 10% of the FDA daily value are bolded.", "steps": [{"name": null, "description": "Values exceeding 10% DV are bolded per FDA guidelines.", "text": "Values exceeding 10% DV are bolded per FDA guidelines.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 8, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather recipe and nutritional information.", "steps": [{"name": null, "description": "Recipe URL (minimalistbaker.com) visited (1 pt).", "text": "Recipe URL (minimalistbaker.com) visited (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for extra-firm tofu (1 pt).", "text": "USDA database URL visited for extra-firm tofu (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for brown rice (1 pt).", "text": "USDA database URL visited for brown rice (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for garlic (1 pt).", "text": "USDA database URL visited for garlic (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for green onion (1 pt).", "text": "USDA database URL visited for green onion (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for peas (1 pt).", "text": "USDA database URL visited for peas (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for carrots (1 pt).", "text": "USDA database URL visited for carrots (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for soy sauce (1 pt).", "text": "USDA database URL visited for soy sauce (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for peanut butter (1 pt).", "text": "USDA database URL visited for peanut butter (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for brown sugar (1 pt).", "text": "USDA database URL visited for brown sugar (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for chili garlic sauce (1 pt).", "text": "USDA database URL visited for chili garlic sauce (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 8, "n_steps": 70, "total_points": 169, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_2_personal_recipe.5", "short_id": "2e", "family": "sheets_2_personal_recipe_foodcomposition", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Recipes", "prompt": "From the recipe https://www.allrecipes.com/recipe/16248/easy-homemade-chili/, pull the average amount of iron, carbohydrates, fat, sugar, potassium, vitamin A, protein, sodium, fiber, calcium, and vitamin C for each ingredient (based on its quantity) using the USDA food database. When searching the USDA, use the food code that best matches the ingredient; more general codes are usually preferable, unless the recipe specifies a matching brand. For instance, if a listed ingredient is just \"Paneer\", you would use \"Cheese, paneer\" instead of \"Palak Paneer\". You may use any of the five USDA food search tabs. Place the results into a Google Sheet as follows. Group nutrients into macros, minerals, and vitamins. Within each group, sort nutrients alphabetically. The first column should be titled \"Ingredients\" and the second \"Link\", followed by sorted nutrient columns starting with macros, then minerals, then vitamins. Above each group, add a merged, centered, italicized header cell labeled \"Macros\", \"Minerals\", or \"Vitamins\" spanning all associated columns. Bold all column titles, excluding the group header cells. Color all cells in each group with a light background color, making sure each group has a distinct color. Place a horizontal line under column titles. List ingredients in the first column in the order they appear in the recipe. In the \"Link\" column, add the USDA data link you used to obtain the quantity of a nutrient in a specific food. Bold any nutrient values exceeding 10% DV according to FDA guidelines. For ingredients you can't find in the USDA database, leave centered N/A. For ingredients with a range of values as the amount, use the minimum value. Ignore ingredients that are listed as optional or don't have a specified amount.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure & Column Layout", "description": "The spreadsheet contains the required columns in the correct order with proper nutrient groupings.", "steps": [{"name": null, "description": "\"Ingredients\" column exists and is first.", "text": "\"Ingredients\" column exists and is first.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Link\" column exists and is second.", "text": "\"Link\" column exists and is second.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Carbohydrates column present.", "text": "Carbohydrates column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fat column present.", "text": "Fat column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Fiber column present.", "text": "Fiber column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Protein column present.", "text": "Protein column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sugar column present.", "text": "Sugar column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Calcium column present.", "text": "Calcium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Iron column present.", "text": "Iron column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Potassium column present.", "text": "Potassium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sodium column present.", "text": "Sodium column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A column present.", "text": "Vitamin A column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C column present.", "text": "Vitamin C column present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Macros columns are in alphabetical order.", "text": "Macros columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Minerals columns are in alphabetical order.", "text": "Minerals columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamins columns are in alphabetical order.", "text": "Vitamins columns are in alphabetical order.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Group Headers & Formatting", "description": "The spreadsheet has properly formatted group headers for nutrient categories.", "steps": [{"name": null, "description": "\"Macros\" merged header exists spanning macro columns.", "text": "\"Macros\" merged header exists spanning macro columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Minerals\" merged header exists spanning mineral columns.", "text": "\"Minerals\" merged header exists spanning mineral columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "\"Vitamins\" merged header exists spanning vitamin columns.", "text": "\"Vitamins\" merged header exists spanning vitamin columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Group headers are centered and italicized.", "text": "Group headers are centered and italicized.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Column titles are bolded (excluding group headers).", "text": "Column titles are bolded (excluding group headers).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Horizontal line under column titles exists.", "text": "Horizontal line under column titles exists.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Color Formatting", "description": "Each nutrient group has a distinct background color applied to all cells in that group.", "steps": [{"name": null, "description": "Macro columns share same background color.", "text": "Macro columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Mineral columns share same background color.", "text": "Mineral columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin columns share same background color.", "text": "Vitamin columns share same background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Three groups have distinct colors from each other.", "text": "Three groups have distinct colors from each other.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Ingredients Present", "description": "The correct ingredients from the recipe are present in the spreadsheet, excluding those without specified amounts.", "steps": [{"name": null, "description": "ground beef row present.", "text": "ground beef row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "onion row present.", "text": "onion row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "tomato sauce row present.", "text": "tomato sauce row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "kidney beans row present.", "text": "kidney beans row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "stewed tomatoes row present.", "text": "stewed tomatoes row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "chili powder row present.", "text": "chili powder row present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "water row NOT present (excluded - no specified amount).", "text": "water row NOT present (excluded - no specified amount).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "USDA Links Validation", "description": "Each ingredient has a valid USDA FoodData Central link that corresponds to the correct food item.", "steps": [{"name": null, "description": "ground beef link valid and matches ingredient (via HTML parsing).", "text": "ground beef link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "onion link valid and matches ingredient (via HTML parsing).", "text": "onion link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "tomato sauce link valid and matches ingredient (via HTML parsing).", "text": "tomato sauce link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "kidney beans link valid and matches ingredient (via HTML parsing).", "text": "kidney beans link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "stewed tomatoes link valid and matches ingredient (via HTML parsing).", "text": "stewed tomatoes link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "chili powder link valid and matches ingredient (via HTML parsing).", "text": "chili powder link valid and matches ingredient (via HTML parsing).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Nutrient Values Accuracy", "description": "Nutrient values for each ingredient match the expected gold label values within tolerance. Each nutrient column is worth 6 points (one per ingredient).", "steps": [{"name": null, "description": "Carbohydrates values match gold labels within tolerance (6 pts).", "text": "Carbohydrates values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fat values match gold labels within tolerance (6 pts).", "text": "Fat values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fiber values match gold labels within tolerance (6 pts).", "text": "Fiber values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Protein values match gold labels within tolerance (6 pts).", "text": "Protein values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sugar values match gold labels within tolerance (6 pts).", "text": "Sugar values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Calcium values match gold labels within tolerance (6 pts).", "text": "Calcium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Iron values match gold labels within tolerance (6 pts).", "text": "Iron values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Potassium values match gold labels within tolerance (6 pts).", "text": "Potassium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Sodium values match gold labels within tolerance (6 pts).", "text": "Sodium values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin A values match gold labels within tolerance (6 pts).", "text": "Vitamin A values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Vitamin C values match gold labels within tolerance (6 pts).", "text": "Vitamin C values match gold labels within tolerance (6 pts).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Bold Formatting for >10% DV", "description": "Nutrient values exceeding 10% of the FDA daily value are bolded.", "steps": [{"name": null, "description": "Values exceeding 10% DV are bolded per FDA guidelines.", "text": "Values exceeding 10% DV are bolded per FDA guidelines.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 8, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather recipe and nutritional information.", "steps": [{"name": null, "description": "Recipe URL (allrecipes.com) visited (1 pt).", "text": "Recipe URL (allrecipes.com) visited (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for ground beef (1 pt).", "text": "USDA database URL visited for ground beef (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for onion (1 pt).", "text": "USDA database URL visited for onion (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for tomato sauce (1 pt).", "text": "USDA database URL visited for tomato sauce (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for kidney beans (1 pt).", "text": "USDA database URL visited for kidney beans (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for stewed tomatoes (1 pt).", "text": "USDA database URL visited for stewed tomatoes (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "USDA database URL visited for chili powder (1 pt).", "text": "USDA database URL visited for chili powder (1 pt).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 8, "n_steps": 58, "total_points": 113, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_2_personal_recipe_foodcomposition/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_38_apartment_finder.1", "short_id": "38a", "family": "sheets_38_apartment_finder", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Housing Search", "prompt": "I am looking to rent a 1 bed 1 bath unit in Salt Lake City, Utah. Your task is to check www.craigslist.org and create a google sheet with at least 5 listings. My budget is 800-1200 USD/month and I would like to move in within the next month. The house should have the features: in-unit laundry and pet-friendly and optionally has at least 450 sq ft. For each house include the necessary columns (including the address) and a link to the listing in the spreadsheet and use conditional formatting on the numeric features (green to red where green means better and red means worse). Add an additional column at the end on interesting positive features of the listing and another column on potential dealbreakers (if there are none, keep it empty). Create a separate \"summary statistics\" table in the top-right corner (starting at column K) while the main data starts at A1. Populate the summary statistics table with data about to listings table using sheets equations that will update the statistics automatically. Ensure all of the text in both tables is visible and not cut off. Both tables must be inserted as blank Google Sheets Tables and then populated with the data.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains columns for all desired features and at least 5 listings.", "steps": [{"name": null, "description": "There is a column for the listing address.", "text": "There is a column for the listing address.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the monthly rent/price.", "text": "There is a column for the monthly rent/price.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bedrooms.", "text": "There is a column for the number of bedrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bathrooms.", "text": "There is a column for the number of bathrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for square footage.", "text": "There is a column for square footage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating in-unit laundry availability.", "text": "There is a column indicating in-unit laundry availability.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating pet-friendliness.", "text": "There is a column indicating pet-friendliness.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column containing a link/URL to each listing.", "text": "There is a column containing a link/URL to each listing.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for interesting positive features.", "text": "There is a column for interesting positive features.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for potential dealbreakers.", "text": "There is a column for potential dealbreakers.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Listing Data Accuracy (x5 listings, 7 pts each)", "description": "Each listing's data in the spreadsheet is verified against the actual Craigslist listing page via HTML parsing and LLM extraction.", "steps": [{"name": null, "description": "The listing URL is valid and accessible.", "text": "The listing URL is valid and accessible.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "text": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "text": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The in-unit laundry status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The in-unit laundry status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The pet-friendly status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The pet-friendly status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Website Visit Validation", "description": "The agent visited Craigslist to find the listings.", "steps": [{"name": null, "description": "The browsing history contains a visit to craigslist.org.", "text": "The browsing history contains a visit to craigslist.org.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The listing URLs in the spreadsheet appear in the browsing history.", "text": "The listing URLs in the spreadsheet appear in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Conditional Formatting Applied", "description": "Conditional formatting is correctly applied to numeric columns.", "steps": [{"name": null, "description": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "text": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "text": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Summary Statistics Table", "description": "A summary statistics table exists with auto-updating formulas.", "steps": [{"name": null, "description": "A summary statistics table exists starting at column K (top-right area of sheet).", "text": "A summary statistics table exists starting at column K (top-right area of sheet).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "text": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Text Visibility and Formatting", "description": "All text in both tables is fully visible and not cut off.", "steps": [{"name": null, "description": "All column headers in the main table are fully visible (not truncated).", "text": "All column headers in the main table are fully visible (not truncated).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All data cells in the main table have adequate column width to display content without truncation.", "text": "All data cells in the main table have adequate column width to display content without truncation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All text in the summary statistics table is fully visible.", "text": "All text in the summary statistics table is fully visible.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text is hidden due to cell overflow issues.", "text": "No text is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 27, "total_points": 55, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_38_apartment_finder.2", "short_id": "38b", "family": "sheets_38_apartment_finder", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Housing Search", "prompt": "I am looking to rent a 2 bed 1 bath apartment in Los Angeles, California. Your task is to check www.craigslist.org and create a google sheet with at least 3 listings. My budget is 2000-3000 USD/month and I would like to move in within the next month. The house should have the features: be fully furnished and optionally has at least 600 sq ft. For each apartment include the necessary columns (including the address) and a link to the listing in the spreadsheet and use conditional formatting on the numeric features (green to red where green means better and red means worse). Add an additional column at the end on interesting positive features of the listing and another column on potential dealbreakers (if there are none, keep it empty). Create a separate \"summary statistics\" table in the top-right corner (starting at column L) while the main data starts at A1. Populate the summary statistics table with data about to listings table using sheets equations that will update the statistics automatically. Ensure all of the text in both tables is visible and not cut off. Both tables must be inserted as blank Google Sheets Tables and then populated with the data.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains columns for all desired features and at least 3 listings.", "steps": [{"name": null, "description": "There is a column for the listing address.", "text": "There is a column for the listing address.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the monthly rent/price.", "text": "There is a column for the monthly rent/price.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bedrooms.", "text": "There is a column for the number of bedrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bathrooms.", "text": "There is a column for the number of bathrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for square footage.", "text": "There is a column for square footage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating if the unit is fully furnished.", "text": "There is a column indicating if the unit is fully furnished.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column containing a link/URL to each listing.", "text": "There is a column containing a link/URL to each listing.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for positive features.", "text": "There is a column for positive features.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for dealbreakers.", "text": "There is a column for dealbreakers.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Listing Data Accuracy (x3 listings, 6 pts each)", "description": "Each listing's data in the spreadsheet is verified against the actual Craigslist listing page via HTML parsing and LLM extraction.", "steps": [{"name": null, "description": "The listing URL is valid and accessible.", "text": "The listing URL is valid and accessible.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "text": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "text": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The fully furnished status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The fully furnished status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Website Visit Validation", "description": "The agent visited Craigslist to find the listings.", "steps": [{"name": null, "description": "The browsing history contains a visit to craigslist.org.", "text": "The browsing history contains a visit to craigslist.org.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The listing URLs in the spreadsheet appear in the browsing history.", "text": "The listing URLs in the spreadsheet appear in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Conditional Formatting Applied", "description": "Conditional formatting is correctly applied to numeric columns.", "steps": [{"name": null, "description": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "text": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "text": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Summary Statistics Table", "description": "A summary statistics table exists with auto-updating formulas.", "steps": [{"name": null, "description": "A summary statistics table exists starting at column L (top-right area of sheet).", "text": "A summary statistics table exists starting at column L (top-right area of sheet).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "text": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Text Visibility and Formatting", "description": "All text in both tables is fully visible and not cut off.", "steps": [{"name": null, "description": "All column headers in the main table are fully visible (not truncated).", "text": "All column headers in the main table are fully visible (not truncated).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All data cells in the main table have adequate column width to display content without truncation.", "text": "All data cells in the main table have adequate column width to display content without truncation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All text in the summary statistics table is fully visible.", "text": "All text in the summary statistics table is fully visible.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text is hidden due to cell overflow issues.", "text": "No text is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 25, "total_points": 37, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_38_apartment_finder.3", "short_id": "38c", "family": "sheets_38_apartment_finder", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Housing Search", "prompt": "I am looking to rent a 4 bed 2 bath house in Houston, TX. Your task is to check www.craigslist.org and create a google sheet with at least 5 listings. My budget is 1200 - 1800 USD/month and I would like to move in within the next month. The house should have the features: no application fee, off-street parking, and optionally has at least 1000 sq ft. For each apartment include the necessary columns (including the address) and a link to the listing in the spreadsheet and use conditional formatting on the numeric features (green to red where green means better and red means worse). Add an additional column at the end on interesting positive features of the listing and another column on potential dealbreakers (if there are none, keep it empty). Create a separate \"summary statistics\" table in the top-right corner (starting at column M) while the main data starts at A1. Populate the summary statistics table with data about the listings table using sheets equations that will update the statistics automatically. Ensure all of the text in both tables is visible and not cut off. Both tables must be inserted as blank Google Sheets Tables and then populated with the data.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains columns for all desired features and at least 5 listings.", "steps": [{"name": null, "description": "There is a column for the listing address.", "text": "There is a column for the listing address.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the monthly rent/price.", "text": "There is a column for the monthly rent/price.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bedrooms.", "text": "There is a column for the number of bedrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bathrooms.", "text": "There is a column for the number of bathrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for square footage.", "text": "There is a column for square footage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating no application fee.", "text": "There is a column indicating no application fee.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating off-street parking availability.", "text": "There is a column indicating off-street parking availability.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column containing a link/URL to each listing.", "text": "There is a column containing a link/URL to each listing.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for positive features.", "text": "There is a column for positive features.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for dealbreakers.", "text": "There is a column for dealbreakers.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Listing Data Accuracy (x5 listings, 7 pts each)", "description": "Each listing's data in the spreadsheet is verified against the actual Craigslist listing page via HTML parsing and LLM extraction.", "steps": [{"name": null, "description": "The listing URL is valid and accessible.", "text": "The listing URL is valid and accessible.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "text": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "text": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The no application fee status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The no application fee status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The off-street parking status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The off-street parking status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Website Visit Validation", "description": "The agent visited Craigslist to find the listings.", "steps": [{"name": null, "description": "The browsing history contains a visit to craigslist.org.", "text": "The browsing history contains a visit to craigslist.org.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The listing URLs in the spreadsheet appear in the browsing history.", "text": "The listing URLs in the spreadsheet appear in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Conditional Formatting Applied", "description": "Conditional formatting is correctly applied to numeric columns.", "steps": [{"name": null, "description": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "text": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "text": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Summary Statistics Table", "description": "A summary statistics table exists with auto-updating formulas.", "steps": [{"name": null, "description": "A summary statistics table exists starting at column M (top-right area of sheet).", "text": "A summary statistics table exists starting at column M (top-right area of sheet).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "text": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Text Visibility and Formatting", "description": "All text in both tables is fully visible and not cut off.", "steps": [{"name": null, "description": "All column headers in the main table are fully visible (not truncated).", "text": "All column headers in the main table are fully visible (not truncated).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All data cells in the main table have adequate column width to display content without truncation.", "text": "All data cells in the main table have adequate column width to display content without truncation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All text in the summary statistics table is fully visible.", "text": "All text in the summary statistics table is fully visible.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text is hidden due to cell overflow issues.", "text": "No text is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 27, "total_points": 55, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_38_apartment_finder.4", "short_id": "38d", "family": "sheets_38_apartment_finder", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Housing Search", "prompt": "I am looking to rent a studio unit in Seattle, Washington. Your task is to check www.craigslist.org and create a google sheet with at least 4 listings. My budget is 1400-1800 USD/month and I would like to move in within the next month. The unit should have the features: fully-furnished, on-site laundry, and optionally has at least 300 sq ft. For each unit include the necessary columns (including the address) and a link to the listing in the spreadsheet and use conditional formatting on the numeric features (green to red where green means better and red means worse). Add an additional column at the end on interesting positive features of the listing and another column on potential dealbreakers (if there are none, keep it empty). Create a separate \"summary statistics\" table in the top-right corner (starting at column M) while the main data starts at A1. Populate the summary statistics table with data about the listings table using sheets equations that will update the statistics automatically. Ensure all of the text in both tables is visible and not cut off. Both tables must be inserted as blank Google Sheets Tables and then populated with the data.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains columns for all desired features and at least 4 listings.", "steps": [{"name": null, "description": "There is a column for the listing address.", "text": "There is a column for the listing address.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the monthly rent/price.", "text": "There is a column for the monthly rent/price.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bedrooms.", "text": "There is a column for the number of bedrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bathrooms.", "text": "There is a column for the number of bathrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for square footage.", "text": "There is a column for square footage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating if the unit is fully furnished.", "text": "There is a column indicating if the unit is fully furnished.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating on-site laundry availability.", "text": "There is a column indicating on-site laundry availability.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column containing a link/URL to each listing.", "text": "There is a column containing a link/URL to each listing.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for positive features.", "text": "There is a column for positive features.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for dealbreakers.", "text": "There is a column for dealbreakers.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Listing Data Accuracy (x4 listings, 7 pts each)", "description": "Each listing's data in the spreadsheet is verified against the actual Craigslist listing page via HTML parsing and LLM extraction.", "steps": [{"name": null, "description": "The listing URL is valid and accessible.", "text": "The listing URL is valid and accessible.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "text": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "text": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The fully furnished status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The fully furnished status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The on-site laundry status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The on-site laundry status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Website Visit Validation", "description": "The agent visited Craigslist to find the listings.", "steps": [{"name": null, "description": "The browsing history contains a visit to craigslist.org.", "text": "The browsing history contains a visit to craigslist.org.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The listing URLs in the spreadsheet appear in the browsing history.", "text": "The listing URLs in the spreadsheet appear in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Conditional Formatting Applied", "description": "Conditional formatting is correctly applied to numeric columns.", "steps": [{"name": null, "description": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "text": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "text": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Summary Statistics Table", "description": "A summary statistics table exists with auto-updating formulas.", "steps": [{"name": null, "description": "A summary statistics table exists starting at column M (top-right area of sheet).", "text": "A summary statistics table exists starting at column M (top-right area of sheet).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "text": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Text Visibility and Formatting", "description": "All text in both tables is fully visible and not cut off.", "steps": [{"name": null, "description": "All column headers in the main table are fully visible (not truncated).", "text": "All column headers in the main table are fully visible (not truncated).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All data cells in the main table have adequate column width to display content without truncation.", "text": "All data cells in the main table have adequate column width to display content without truncation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All text in the summary statistics table is fully visible.", "text": "All text in the summary statistics table is fully visible.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text is hidden due to cell overflow issues.", "text": "No text is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 27, "total_points": 48, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_38_apartment_finder.5", "short_id": "38e", "family": "sheets_38_apartment_finder", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Housing Search", "prompt": "I am looking to rent a 3 bed 2 bath unit in Chicago, Illinois. Your task is to check www.craigslist.org and create a google sheet with at least 6 listings. My budget is 3000-4500 USD/month and I would like to move in within the next month. The unit should have the features: air-conditioning and optionally has at least 1200 sq ft. For each unit include the necessary columns (including the address) and a link to the listing in the spreadsheet and use conditional formatting on the numeric features (green to red where green means better and red means worse). Add an additional column at the end on interesting positive features of the listing and another column on potential dealbreakers (if there are none, keep it empty). Create a separate \"summary statistics\" table in the top-right corner (starting at column L) while the main data starts at A1. Populate the summary statistics table with data about the listings table using sheets equations that will update the statistics automatically. Ensure all of the text in both tables is visible and not cut off. Both tables must be inserted as blank Google Sheets Tables and then populated with the data.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains columns for all desired features and at least 6 listings.", "steps": [{"name": null, "description": "There is a column for the listing address.", "text": "There is a column for the listing address.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the monthly rent/price.", "text": "There is a column for the monthly rent/price.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bedrooms.", "text": "There is a column for the number of bedrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for the number of bathrooms.", "text": "There is a column for the number of bathrooms.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for square footage.", "text": "There is a column for square footage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column indicating air conditioning availability.", "text": "There is a column indicating air conditioning availability.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column containing a link/URL to each listing.", "text": "There is a column containing a link/URL to each listing.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for positive features.", "text": "There is a column for positive features.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column for dealbreakers.", "text": "There is a column for dealbreakers.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Listing Data Accuracy (x6 listings, 6 pts each)", "description": "Each listing's data in the spreadsheet is verified against the actual Craigslist listing page via HTML parsing and LLM extraction.", "steps": [{"name": null, "description": "The listing URL is valid and accessible.", "text": "The listing URL is valid and accessible.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "text": "The price in the spreadsheet matches the price extracted from the listing HTML (within 5% tolerance).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bedroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "text": "The bathroom count in the spreadsheet matches the value extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "text": "The address in the spreadsheet matches or is contained in the address extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The air conditioning status in the spreadsheet matches the information extracted from the listing HTML.", "text": "The air conditioning status in the spreadsheet matches the information extracted from the listing HTML.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Website Visit Validation", "description": "The agent visited Craigslist to find the listings.", "steps": [{"name": null, "description": "The browsing history contains a visit to craigslist.org.", "text": "The browsing history contains a visit to craigslist.org.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The listing URLs in the spreadsheet appear in the browsing history.", "text": "The listing URLs in the spreadsheet appear in the browsing history.", "category": "web_visit_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Conditional Formatting Applied", "description": "Conditional formatting is correctly applied to numeric columns.", "steps": [{"name": null, "description": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "text": "Conditional formatting (color scale) is applied to at least one numeric column (e.g., price, sq ft).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "text": "The formatting uses a green-to-red scale where green indicates better values and red indicates worse values.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Summary Statistics Table", "description": "A summary statistics table exists with auto-updating formulas.", "steps": [{"name": null, "description": "A summary statistics table exists starting at column L (top-right area of sheet).", "text": "A summary statistics table exists starting at column L (top-right area of sheet).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "text": "The summary table contains formulas/equations that reference the main listing data (not hardcoded values).", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Text Visibility and Formatting", "description": "All text in both tables is fully visible and not cut off.", "steps": [{"name": null, "description": "All column headers in the main table are fully visible (not truncated).", "text": "All column headers in the main table are fully visible (not truncated).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All data cells in the main table have adequate column width to display content without truncation.", "text": "All data cells in the main table have adequate column width to display content without truncation.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "All text in the summary statistics table is fully visible.", "text": "All text in the summary statistics table is fully visible.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text is hidden due to cell overflow issues.", "text": "No text is hidden due to cell overflow issues.", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 25, "total_points": 55, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_38_apartment_finder/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_45_wedding_planner.1", "short_id": "24a", "family": "sheets_45_Personal_WeddingPlanner_weddingcolorpallette", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Wedding Planner", "prompt": "I want to make a Google Sheet to help choose color templates for my wedding. First, do a web search and find at least 2 articles on wedding colors. From those articles, extract at least 10 distinct shades of blue, green, or red. Then, at the top left of the sheet, list all the colors vertically. In the column to the right of the color name, put the link to the article that mentioned the color. Then, in the next column to the right, fill in that cell with a background color matching the named color as closely as possible using the color's standard hex code (e.g. from colorhexa.com). Then, in the next column to the right, find a page or link to that specific color at a paint store and paste the link. Below this list, on the same sheet, add a second table with the same colors as the list above as column headers, in the same order. The rows are different wedding objects with at least 3 objects. Fill in each cell with a real photograph from the internet of that decoration in that color and display it in using =IMAGE() in the sheet. If you can't find an image for an object in a specific color, skip it, but fill in at least half of the cells. Finally, in a new tab, create at least 10 different rows of full-color palettes. For each row, choose three colors from the list above and fill each cell's background color with one of those three colors to show different color palettes.", "checkpoints": [{"n": 1, "title": "Color Extraction", "description": "The agent found and extracted wedding color information from at least two articles and color names are listed vertically in the top-left area of the main sheet.", "steps": [{"name": null, "description": "All extracted color names appear in a vertical list in a single column.", "text": "All extracted color names appear in a vertical list in a single column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "text": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 unique color names/shades are present, extracted from articles.", "text": "At least 10 unique color names/shades are present, extracted from articles.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The agent searched for and found at least 2 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "text": "The agent searched for and found at least 2 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "All extracted colors belong to the specified blue, green, or red categories (LLM judge that each color falls into the category).", "text": "All extracted colors belong to the specified blue, green, or red categories (LLM judge that each color falls into the category).", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 2, "title": "Article Source Links", "description": "Article links are placed in the column immediately to the right of each color name.", "steps": [{"name": null, "description": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color name has a corresponding article link in the same row. (1 pt)", "text": "Each color name has a corresponding article link in the same row. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The article links are functional and reachable. (1 pt)", "text": "The article links are functional and reachable. (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "text": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Color Cell Formatting", "description": "Cells are filled with colors matching the color names in the third column.", "steps": [{"name": null, "description": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "text": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "text": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "text": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each color name has a corresponding colored cell. (1 pt)", "text": "Each color name has a corresponding colored cell. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Paint Store References", "description": "Paint store links are provided in the fourth column next to each color.", "steps": [{"name": null, "description": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color has an associated paint store link. (1 pt)", "text": "Each color has an associated paint store link. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "text": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Wedding Decoration Matrix", "description": "A wedding decoration matrix is created below the color list with images.", "steps": [{"name": null, "description": "At least 3 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "text": "At least 3 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "text": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matrix column headers appear in the same order as the original color list. (1 pt)", "text": "The matrix column headers appear in the same order as the original color list. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least half of the matrix cells contain images. (2 pts)", "text": "At least half of the matrix cells contain images. (2 pts)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "text": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Color Palette Tab", "description": "A separate tab contains at least 10 color palette combinations.", "steps": [{"name": null, "description": "A new sheet/tab was created for color palettes.", "text": "A new sheet/tab was created for color palettes.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 rows of color combinations exist.", "text": "At least 10 rows of color combinations exist.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row contains exactly 3 colored cells representing a palette.", "text": "Each row contains exactly 3 colored cells representing a palette.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Colors are filled as background colors (not just text).", "text": "Colors are filled as background colors (not just text).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Color combinations use colors from the original extracted list.", "text": "Color combinations use colors from the original extracted list.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 26, "total_points": 63, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_45_wedding_planner.2", "short_id": "24b", "family": "sheets_45_Personal_WeddingPlanner_weddingcolorpallette", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Wedding Planner", "prompt": "I want to make a Google Sheet to help choose color templates for my wedding. First, do a web search and find at least 3 articles on wedding colors. From those articles, extract at least 10 distinct shades of purple, yellow, or orange. Then, at the top left of the sheet, list all the colors vertically. In the column to the right of the color name, put the link to the article that mentioned the color. Then, in the next column to the right, fill in that cell with a background color matching the named color as closely as possible using the color's standard hex code (e.g. from colorhexa.com). Then, in the next column to the right, find a page or link to that specific color at a paint store and paste the link. Below this list, on the same sheet, add a second table with the same colors as the list above as column headers, in the same order. The rows are different wedding objects with at least 3 objects. Fill in each cell with a real photograph from the internet of that decoration in that color and display it in using =IMAGE() in the sheet. If you can't find an image for an object in a specific color, skip it, but fill in at least half of the cells. Finally, in a new tab, create at least 10 different rows of full-color palettes. For each row, choose three colors from the list above and fill each cell's background color with one of those three colors to show different color palettes.", "checkpoints": [{"n": 1, "title": "Color Extraction", "description": "The agent found and extracted wedding color information from at least three articles and color names are listed vertically in the top-left area of the main sheet.", "steps": [{"name": null, "description": "All extracted color names appear in a vertical list in a single column.", "text": "All extracted color names appear in a vertical list in a single column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "text": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 unique color names/shades are present, extracted from articles.", "text": "At least 10 unique color names/shades are present, extracted from articles.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The agent searched for and found at least 3 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "text": "The agent searched for and found at least 3 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "All extracted colors belong to the specified purple, yellow, or orange categories (LLM judge that each color falls into the category).", "text": "All extracted colors belong to the specified purple, yellow, or orange categories (LLM judge that each color falls into the category).", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 2, "title": "Article Source Links", "description": "Article links are placed in the column immediately to the right of each color name.", "steps": [{"name": null, "description": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color name has a corresponding article link in the same row. (1 pt)", "text": "Each color name has a corresponding article link in the same row. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The article links are functional and reachable. (1 pt)", "text": "The article links are functional and reachable. (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "text": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Color Cell Formatting", "description": "Cells are filled with colors matching the color names in the third column.", "steps": [{"name": null, "description": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "text": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "text": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "text": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each color name has a corresponding colored cell. (1 pt)", "text": "Each color name has a corresponding colored cell. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Paint Store References", "description": "Paint store links are provided in the fourth column next to each color.", "steps": [{"name": null, "description": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color has an associated paint store link. (1 pt)", "text": "Each color has an associated paint store link. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "text": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Wedding Decoration Matrix", "description": "A wedding decoration matrix is created below the color list with images.", "steps": [{"name": null, "description": "At least 3 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "text": "At least 3 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "text": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matrix column headers appear in the same order as the original color list. (1 pt)", "text": "The matrix column headers appear in the same order as the original color list. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least half of the matrix cells contain images. (2 pts)", "text": "At least half of the matrix cells contain images. (2 pts)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "text": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Color Palette Tab", "description": "A separate tab contains at least 10 color palette combinations.", "steps": [{"name": null, "description": "A new sheet/tab was created for color palettes.", "text": "A new sheet/tab was created for color palettes.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 rows of color combinations exist.", "text": "At least 10 rows of color combinations exist.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row contains exactly 3 colored cells representing a palette.", "text": "Each row contains exactly 3 colored cells representing a palette.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Colors are filled as background colors (not just text).", "text": "Colors are filled as background colors (not just text).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Color combinations use colors from the original extracted list.", "text": "Color combinations use colors from the original extracted list.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 26, "total_points": 63, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_45_wedding_planner.3", "short_id": "24c", "family": "sheets_45_Personal_WeddingPlanner_weddingcolorpallette", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Wedding Planner", "prompt": "I want to make a Google Sheet to help choose color templates for my wedding. First, do a web search and find at least 2 articles on wedding colors. From those articles, extract at least 10 distinct shades of ivory, champagne, or blush. Then, at the top left of the sheet, list all the colors vertically. In the column to the right of the color name, put the link to the article that mentioned the color. Then, in the next column to the right, fill in that cell with a background color matching the named color as closely as possible using the color's standard hex code (e.g. from colorhexa.com). Then, in the next column to the right, find a page or link to that specific color at a paint store and paste the link. Below this list, on the same sheet, add a second table with the same colors as the list above as column headers, in the same order. The rows are different wedding objects with at least 5 objects. Fill in each cell with a real photograph from the internet of that decoration in that color and display it in using =IMAGE() in the sheet. If you can't find an image for an object in a specific color, skip it, but fill in at least half of the cells. Finally, in a new tab, create at least 10 different rows of full-color palettes. For each row, choose four colors from the list above and fill each cell's background color with one of those four colors to show different color palettes.", "checkpoints": [{"n": 1, "title": "Color Extraction", "description": "The agent found and extracted wedding color information from at least two articles and color names are listed vertically in the top-left area of the main sheet.", "steps": [{"name": null, "description": "All extracted color names appear in a vertical list in a single column.", "text": "All extracted color names appear in a vertical list in a single column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "text": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 unique color names/shades are present, extracted from articles.", "text": "At least 10 unique color names/shades are present, extracted from articles.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The agent searched for and found at least 2 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "text": "The agent searched for and found at least 2 articles about wedding colors (agent trace, LLM judge that the articles are about weddings).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "All extracted colors belong to the specified ivory, champagne, or blush categories (LLM judge that each color falls into the category).", "text": "All extracted colors belong to the specified ivory, champagne, or blush categories (LLM judge that each color falls into the category).", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 2, "title": "Article Source Links", "description": "Article links are placed in the column immediately to the right of each color name.", "steps": [{"name": null, "description": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color name has a corresponding article link in the same row. (1 pt)", "text": "Each color name has a corresponding article link in the same row. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The article links are functional and reachable. (1 pt)", "text": "The article links are functional and reachable. (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "text": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Color Cell Formatting", "description": "Cells are filled with colors matching the color names in the third column.", "steps": [{"name": null, "description": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "text": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "text": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "text": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each color name has a corresponding colored cell. (1 pt)", "text": "Each color name has a corresponding colored cell. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Paint Store References", "description": "Paint store links are provided in the fourth column next to each color.", "steps": [{"name": null, "description": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color has an associated paint store link. (1 pt)", "text": "Each color has an associated paint store link. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "text": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Wedding Decoration Matrix", "description": "A wedding decoration matrix is created below the color list with images.", "steps": [{"name": null, "description": "At least 5 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "text": "At least 5 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "text": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matrix column headers appear in the same order as the original color list. (1 pt)", "text": "The matrix column headers appear in the same order as the original color list. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least half of the matrix cells contain images. (2 pts)", "text": "At least half of the matrix cells contain images. (2 pts)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "text": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Color Palette Tab", "description": "A separate tab contains at least 10 four-color palette combinations.", "steps": [{"name": null, "description": "A new sheet/tab was created for color palettes.", "text": "A new sheet/tab was created for color palettes.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 rows of color combinations exist.", "text": "At least 10 rows of color combinations exist.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row contains exactly 4 colored cells representing a palette.", "text": "Each row contains exactly 4 colored cells representing a palette.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Colors are filled as background colors (not just text).", "text": "Colors are filled as background colors (not just text).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Color combinations use colors from the original extracted list.", "text": "Color combinations use colors from the original extracted list.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 6, "n_steps": 26, "total_points": 63, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_45_wedding_planner.4", "short_id": "24d", "family": "sheets_45_Personal_WeddingPlanner_weddingcolorpallette", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Wedding Planner", "prompt": "I want to make a Google Sheet to help choose color templates for my fall wedding. First, do a web search and find at least three articles discussing fall or autumn wedding colors. From those articles, extract at least 15 different specific shades of burgundy, rust, or mustard. Then, at the top left of the sheet, list all the colors vertically. In the column to the right of the color name, put the link to the article that mentioned the color. Then, in the next column to the right, fill in that cell with a background color matching the named color as closely as possible using the color's standard hex code (e.g. from colorhexa.com). Then, in the next column to the right, find a page or link to that specific color at a paint store and paste the link. Then, in the next column, add a short one-sentence description of the mood or feeling that color evokes. Below this list, on the same sheet, add a second table with the same colors as the list above as column headers, in the same order. The rows are different wedding objects with at least 4 objects. Fill in each cell with a real photograph from the internet of that decoration in that color and display it in using =IMAGE() in the sheet. If you can't find an image for an object in a specific color, skip it, but fill in at least half of the cells. Finally, in a new tab, create at least 15 different rows of full-color palettes. For each row, choose three colors from the list above and fill each cell's background color with one of those three colors to show different color palettes. Add a label in the first column of each row giving the palette a creative name.", "checkpoints": [{"n": 1, "title": "Color Extraction", "description": "The agent found and extracted fall/autumn wedding color information from at least three articles and color names are listed vertically in the top-left area of the main sheet.", "steps": [{"name": null, "description": "All extracted color names appear in a vertical list in a single column.", "text": "All extracted color names appear in a vertical list in a single column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The color list starts in the top-left area of the sheet (rows 1-20 approximately).", "text": "The color list starts in the top-left area of the sheet (rows 1-20 approximately).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 15 unique color names/shades are present, extracted from articles.", "text": "At least 15 unique color names/shades are present, extracted from articles.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The agent searched for and found at least 3 articles about fall or autumn wedding colors (agent trace, LLM judge that the articles are about fall/autumn weddings).", "text": "The agent searched for and found at least 3 articles about fall or autumn wedding colors (agent trace, LLM judge that the articles are about fall/autumn weddings).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "All extracted colors belong to the specified burgundy, rust, or mustard categories (LLM judge that each color falls into the category).", "text": "All extracted colors belong to the specified burgundy, rust, or mustard categories (LLM judge that each color falls into the category).", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 2, "title": "Article Source Links", "description": "Article links are placed in the column immediately to the right of each color name.", "steps": [{"name": null, "description": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color name has a corresponding article link in the same row. (1 pt)", "text": "Each color name has a corresponding article link in the same row. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The article links are functional and reachable. (1 pt)", "text": "The article links are functional and reachable. (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "text": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Color Cell Formatting", "description": "Cells are filled with colors matching the color names in the third column.", "steps": [{"name": null, "description": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "text": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "text": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "text": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each color name has a corresponding colored cell. (1 pt)", "text": "Each color name has a corresponding colored cell. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Paint Store References", "description": "Paint store links are provided in the fourth column next to each color.", "steps": [{"name": null, "description": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color has an associated paint store link. (1 pt)", "text": "Each color has an associated paint store link. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "text": "The paint store links are functional and reachable. (1 pt) The links lead to relevant paint store / color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Mood Descriptions", "description": "A fifth column contains a one-sentence description of the mood or feeling each color evokes.", "steps": [{"name": null, "description": "Each color has a corresponding mood description (no missing entries). (1 pt)", "text": "Each color has a corresponding mood description (no missing entries). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each description is a short, single sentence (LLM judge). (5 pts, proportional)", "text": "Each description is a short, single sentence (LLM judge). (5 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Descriptions describe the mood or feeling evoked by the color (LLM judge that content is about mood/feeling). (5 pts, proportional)", "text": "Descriptions describe the mood or feeling evoked by the color (LLM judge that content is about mood/feeling). (5 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Mood descriptions are unique across colours (no copy-pasted duplicates). (1 pt)", "text": "Mood descriptions are unique across colours (no copy-pasted duplicates). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Wedding Decoration Matrix", "description": "A wedding decoration matrix is created below the color list with images.", "steps": [{"name": null, "description": "At least 4 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "text": "At least 4 types of wedding decorations are listed in the leftmost column (API for location, LLM judge for content). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "text": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matrix column headers appear in the same order as the original color list. (1 pt)", "text": "The matrix column headers appear in the same order as the original color list. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least half of the matrix cells contain images. (2 pts)", "text": "At least half of the matrix cells contain images. (2 pts)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "text": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "Named Color Palette Tab", "description": "A separate tab contains at least 15 named color palette combinations.", "steps": [{"name": null, "description": "A new sheet/tab was created for color palettes.", "text": "A new sheet/tab was created for color palettes.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 15 rows of color combinations exist.", "text": "At least 15 rows of color combinations exist.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The first column of each row contains a creative palette name (LLM judge that the label is a creative name).", "text": "The first column of each row contains a creative palette name (LLM judge that the label is a creative name).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each row contains exactly 3 colored cells representing the palette.", "text": "Each row contains exactly 3 colored cells representing the palette.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Colors are filled as background colors (not just text).", "text": "Colors are filled as background colors (not just text).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Color combinations use colors from the original extracted list.", "text": "Color combinations use colors from the original extracted list.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 31, "total_points": 76, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_45_wedding_planner.5", "short_id": "24e", "family": "sheets_45_Personal_WeddingPlanner_weddingcolorpallette", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Wedding Planner", "prompt": "I want to make a Google Sheet to help choose color templates for my beach wedding. First, do a web search and find at least two articles discussing beach or coastal wedding colors. From those articles, extract at least 10 different specific shades of coral, turquoise, or sand. Then, at the top left of the sheet, list all the colors vertically. In the column to the right of the color name, put the link to the article that mentioned the color. Then, in the next column to the right, fill in that cell with a background color matching the named color as closely as possible using the color's standard hex code (e.g. from colorhexa.com). Then, in the next column to the right, find a page or link to that specific color at a fabric or textile store (not a paint store) and paste the link. Below this list, on the same sheet, add a second table with the same colors as the list above as column headers, in the same order. The rows are different wedding objects specific to a beach wedding (such as centerpieces, bridesmaid dresses, table runners, shell decorations, and invitations), with at least five objects. Fill in each cell with a real photograph from the internet of that decoration in that color and display it in using =IMAGE() in the sheet. If you can't find an image for an object in a specific color, skip it, but fill in at least half of the cells. Finally, create two new tabs: in the first, create at least 10 rows of three-color palettes. For each row, choose three colors from the list above and fill each cell's background color with one of those three colors to show different color palettes. Label each palette row in the first column as either \"Day Ceremony\" or \"Evening Reception\". In the second new tab, create a simple checklist of wedding decoration items in column A, with columns B through D representing your top three color palette choices, so you can plan which color goes with which item.", "checkpoints": [{"n": 1, "title": "Color Extraction", "description": "The agent found and extracted beach/coastal wedding color information from at least two articles and color names are listed vertically in the top-left area of the main sheet.", "steps": [{"name": null, "description": "All extracted color names appear in a vertical list in a single column.", "text": "All extracted color names appear in a vertical list in a single column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "text": "The color list starts in the top-left area of the sheet (rows 1-15 approximately).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 unique color names/shades are present, extracted from articles.", "text": "At least 10 unique color names/shades are present, extracted from articles.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The agent searched for and found at least 2 articles about beach or coastal wedding colors (agent trace, LLM judge that the articles are about beach/coastal weddings).", "text": "The agent searched for and found at least 2 articles about beach or coastal wedding colors (agent trace, LLM judge that the articles are about beach/coastal weddings).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "All extracted colors belong to the specified coral, turquoise, or sand categories (LLM judge that each color falls into the category).", "text": "All extracted colors belong to the specified coral, turquoise, or sand categories (LLM judge that each color falls into the category).", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 2, "title": "Article Source Links", "description": "Article links are placed in the column immediately to the right of each color name.", "steps": [{"name": null, "description": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The article column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color name has a corresponding article link in the same row. (1 pt)", "text": "Each color name has a corresponding article link in the same row. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The article links are functional and reachable. (1 pt)", "text": "The article links are functional and reachable. (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "text": "The links lead to relevant wedding-color content for that color. (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Color Cell Formatting", "description": "Cells are filled with colors matching the color names in the third column.", "steps": [{"name": null, "description": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "text": "The fill column is a colour-fill column (every populated cell has a background fill, not just typed text). (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "text": "The fill colors match the named colors or are extremely close to them (VLM judge). (5 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "text": "The hex value for the cell color matches the color shade hex found in https://www.colorhexa.com/. (5 pts, proportional)", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Each color name has a corresponding colored cell. (1 pt)", "text": "Each color name has a corresponding colored cell. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Fabric/Textile Store References", "description": "Fabric or textile store links are provided in the fourth column next to each color.", "steps": [{"name": null, "description": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "text": "The store column is a URL/link column (every populated cell contains a URL, not free text). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each color has an associated fabric/textile store link. (1 pt)", "text": "Each color has an associated fabric/textile store link. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The fabric/textile store links are functional and reachable. (1 pt) The links lead to relevant fabric/textile color content for that color (LLM judge that the linked page is from a fabric or textile store, not a paint store). (10 pts, proportional)", "text": "The fabric/textile store links are functional and reachable. (1 pt) The links lead to relevant fabric/textile color content for that color (LLM judge that the linked page is from a fabric or textile store, not a paint store). (10 pts, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Beach Wedding Decoration Matrix", "description": "A beach wedding decoration matrix is created below the color list with images.", "steps": [{"name": null, "description": "At least 5 types of beach wedding decorations are listed in the leftmost column (API for location, LLM judge that items are beach-wedding-specific such as centerpieces, bridesmaid dresses, table runners, shell decorations, or invitations). (1 pt)", "text": "At least 5 types of beach wedding decorations are listed in the leftmost column (API for location, LLM judge that items are beach-wedding-specific such as centerpieces, bridesmaid dresses, table runners, shell decorations, or invitations). (1 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "text": "Column headers contain the same color names from the original list (exact text match). (1 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matrix column headers appear in the same order as the original color list. (1 pt)", "text": "The matrix column headers appear in the same order as the original color list. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least half of the matrix cells contain images. (2 pts)", "text": "At least half of the matrix cells contain images. (2 pts)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "text": "Images show the specified decoration type in the corresponding color (VLM judge). (10 pts, proportional)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Day/Evening Color Palette Tab", "description": "A new tab contains at least 10 labeled color palette combinations.", "steps": [{"name": null, "description": "A new sheet/tab was created for color palettes.", "text": "A new sheet/tab was created for color palettes.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "At least 10 rows of three-color combinations exist.", "text": "At least 10 rows of three-color combinations exist.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row contains exactly 3 colored cells representing a palette.", "text": "Each row contains exactly 3 colored cells representing a palette.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Colors are filled as background colors (not just text).", "text": "Colors are filled as background colors (not just text).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each palette row is labeled either \"Day Ceremony\" or \"Evening Reception\" (exact text match in a label column).", "text": "Each palette row is labeled either \"Day Ceremony\" or \"Evening Reception\" (exact text match in a label column).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Color combinations use colors from the original extracted list.", "text": "Color combinations use colors from the original extracted list.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 7, "title": "Decoration-Palette Checklist Tab", "description": "A second new tab contains a checklist mapping decoration items to top three palette choices.", "steps": [{"name": null, "description": "A second new sheet/tab was created.", "text": "A second new sheet/tab was created.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Column A contains a list of wedding decoration items (LLM judge for content).", "text": "Column A contains a list of wedding decoration items (LLM judge for content).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Columns B, C, and D represent three palette choices (e.g., labeled or color-coded as the top three palettes from the palette tab).", "text": "Columns B, C, and D represent three palette choices (e.g., labeled or color-coded as the top three palettes from the palette tab).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Cells in columns B-D indicate which color from each palette goes with each decoration item (filled with colors or color references).", "text": "Cells in columns B-D indicate which color from each palette goes with each decoration item (filled with colors or color references).", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 31, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_45_Personal_WeddingPlanner_weddingcolorpallette/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_55_movie_recommendation.1", "short_id": null, "family": "sheets_55_Movie_Recommendation", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Media Recommendation", "prompt": "I want to watch a movie tonight. I prefer the following genres: Action, Drama, Thriller, Sci-Fi, and Comedy. Find the genre based on the IMDb listing of the movies.\nI want to watch award-winning movies, and I prefer movies that are highly rated and well-reviewed (IMDb ≥ 6.5). Look for movies that have won at least one Oscar award for Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, or Best Cinematography.\nPlease make a Google Sheet with at least 5 movie recommendations that won at least one of the mentioned awards. Each movie's primary genre on IMDb — the first standard genre IMDb lists for the movie — must be one of the above listed genres. Use IMDb's standard genre list, not UI tags.\nEach row should contain information about one movie. Use these column header names, in this order:\nMovie Title\nGenre\nMPA/Age Rating\nIMDb Score\nRelease Year\nDuration\nOscar Awards Won\n\nMPA/Age Rating is the MPA/age rating, e.g., PG, PG-13, R.\n\nIn the Oscar Awards Won column, list all of the qualifying Oscar categories above that the movie won, separated by commas. Do not include other awards.\nSort the list based on Duration in descending order.\nHighlight the highest IMDb score in the list with green fill + bold text.\nHighlight the lowest IMDb score with red fill + bold text.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns, at least 5 unique movies, and no blank cells.", "steps": [{"name": null, "description": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "text": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "text": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "text": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "text": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No cell in any of the required columns is left blank. (1 pt)", "text": "No cell in any of the required columns is left blank. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Accuracy", "description": "Each movie's IMDb page must appear in the browsing history, and its genre, Oscar wins, IMDb score, and MPA rating are verified against authoritative sources. Each step is scored proportionally per movie: round((movies passing / total movies) * 10). Movies whose IMDb page wasn't visited fail every subsequent step.", "steps": [{"name": null, "description": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "text": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Action, Drama, Thriller, Sci-Fi, Comedy). Word-boundary match: compound forms like \"Dark Comedy\" or \"Romance Comedy\" pass for \"Comedy\"; \"Farce\" does not. (10 pt, proportional)", "text": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Action, Drama, Thriller, Sci-Fi, Comedy). Word-boundary match: compound forms like \"Dark Comedy\" or \"Romance Comedy\" pass for \"Comedy\"; \"Farce\" does not. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "text": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "text": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "text": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sorting and Conditional Formatting", "description": "The list is sorted by duration and the highest/lowest IMDb scores are visually highlighted.", "steps": [{"name": null, "description": "The movie list is sorted by Duration in descending order. (1 pt)", "text": "The movie list is sorted by Duration in descending order. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "text": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "text": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 13, "total_points": 58, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_55_movie_recommendation.2", "short_id": "55b", "family": "sheets_55_Movie_Recommendation", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Media Recommendation", "prompt": "I want to watch a movie tonight. I prefer the following genres: Comedy, Musical, Adventure. Find the genre based on the IMDb listing of the movies.\nI want to watch award-winning movies, released between 1940 and 1960, and I prefer movies that are highly rated and well-reviewed (IMDb ≥ 6.5). Look for movies that have won at least one Oscar award for Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, or Best Cinematography.\nPlease make a Google Sheet with at least 5 movie recommendations that won at least one of the mentioned awards. Each movie's primary genre on IMDb — the first standard genre IMDb lists for the movie — must be one of the above listed genres. Use IMDb's standard genre list, not UI tags.\nEach row should contain information about one movie. Use these column header names, in this order:\nMovie Title\nGenre\nMPA/Age Rating\nIMDb Score\nRelease Year\nDuration\nOscar Awards Won\n\nMPA/Age Rating is the MPA/age rating, e.g., PG, PG-13, R.\n\nIn the Oscar Awards Won column, list all of the qualifying Oscar categories above that the movie won, separated by commas. Do not include other awards.\nSort the list based on Duration in descending order.\nHighlight the highest IMDb score in the list with green fill + bold text.\nHighlight the lowest IMDb score with red fill + bold text.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns, at least 5 unique movies, and no blank cells.", "steps": [{"name": null, "description": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "text": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "text": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "text": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "text": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No cell in any of the required columns is left blank. (1 pt)", "text": "No cell in any of the required columns is left blank. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Accuracy", "description": "Each movie's genre, release time, Oscar wins, IMDb score, and MPA rating are verified against authoritative sources. Each step is scored proportionally per movie: round((movies passing / total movies) * 10).", "steps": [{"name": null, "description": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "text": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Comedy, Musical, Adventure). Word-boundary match: compound forms like \"Dark Comedy\" or \"Romance Comedy\" pass for \"Comedy\"; \"Farce\" does not. (10 pt, proportional)", "text": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Comedy, Musical, Adventure). Word-boundary match: compound forms like \"Dark Comedy\" or \"Romance Comedy\" pass for \"Comedy\"; \"Farce\" does not. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each movie's release year must be between 1940 and 1960. (10 pt, proportional)", "text": "Each movie's release year must be between 1940 and 1960. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "text": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Director, Best Original Screenplay, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "text": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "text": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sorting and Conditional Formatting", "description": "The list is sorted by duration and the highest/lowest IMDb scores are visually highlighted.", "steps": [{"name": null, "description": "The movie list is sorted by Duration in descending order. (1 pt)", "text": "The movie list is sorted by Duration in descending order. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "text": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "text": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 14, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_55_movie_recommendation.3", "short_id": "55c", "family": "sheets_55_Movie_Recommendation", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Media Recommendation", "prompt": "I want to watch a movie tonight. I prefer the following genres: Drama, Western. Find the genre based on the IMDb listing of the movies.\nI want to watch award-winning movies, released between 1950 and 1972, and I prefer movies that are highly rated and well-reviewed (IMDb ≥ 6.5). Look for movies that have won at least one Oscar award for Best Actor in a Leading Role or Best Actor in a Supporting Role.\nPlease make a Google Sheet with at least 5 movie recommendations that won at least one of the mentioned awards. Each movie's primary genre on IMDb — the first standard genre IMDb lists for the movie — must be one of the above listed genres. Use IMDb's standard genre list, not UI tags.\nEach row should contain information about one movie. Use these column header names, in this order:\nMovie Title\nGenre\nMPA/Age Rating\nIMDb Score\nRelease Year\nDuration\nOscar Awards Won\n\nMPA/Age Rating is the MPA/age rating, e.g., PG, PG-13, R.\n\nIn the Oscar Awards Won column, list all of the qualifying Oscar categories above that the movie won, separated by commas. Do not include other awards.\nSort the list based on Duration in descending order.\nHighlight the highest IMDb score in the list with green fill + bold text.\nHighlight the lowest IMDb score with red fill + bold text.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns, at least 5 unique movies, and no blank cells.", "steps": [{"name": null, "description": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "text": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "text": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "text": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "text": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No cell in any of the required columns is left blank. (1 pt)", "text": "No cell in any of the required columns is left blank. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Accuracy", "description": "Each movie's genre, Oscar wins, IMDb score, and MPA rating are verified against authoritative sources. Each step is scored proportionally per movie: round((movies passing / total movies) * 10).", "steps": [{"name": null, "description": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "text": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Western). Word-boundary match: compound forms like \"Dark Drama\" pass for \"Drama\"; unrelated tags do not. (10 pt, proportional)", "text": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Western). Word-boundary match: compound forms like \"Dark Drama\" pass for \"Drama\"; unrelated tags do not. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each movie's release year must be between 1950 and 1972. (10 pt, proportional)", "text": "Each movie's release year must be between 1950 and 1972. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor in a Leading Role, Best Actor in a Supporting Role}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "text": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor in a Leading Role, Best Actor in a Supporting Role}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "text": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "text": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sorting and Conditional Formatting", "description": "The list is sorted by duration and the highest/lowest IMDb scores are visually highlighted.", "steps": [{"name": null, "description": "The movie list is sorted by Duration in descending order. (1 pt)", "text": "The movie list is sorted by Duration in descending order. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "text": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "text": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 14, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_55_movie_recommendation.4", "short_id": "55d", "family": "sheets_55_Movie_Recommendation", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Media Recommendation", "prompt": "I want to watch a movie tonight. I prefer the following genres: Drama, Crime, Thriller. Find the genre based on the IMDb listing of the movies.\nI want to watch award-winning movies, released between 1970 and 2010, and I prefer movies that are highly rated and well-reviewed (IMDb ≥ 6.5). Look for movies that have won at least one Oscar award for Best Actor, Best Actress, Best Adapted Screenplay, or Best Cinematography.\nPlease make a Google Sheet with at least 5 movie recommendations that won at least one of the mentioned awards. Each movie's primary genre on IMDb — the first standard genre IMDb lists for the movie — must be one of the above listed genres. Use IMDb's standard genre list, not UI tags. Release years must be at least 8 years apart from one another.\nEach row should contain information about one movie. Use these column header names, in this order:\nMovie Title\nGenre\nMPA/Age Rating\nIMDb Score\nRelease Year\nDuration\nOscar Awards Won\n\nMPA/Age Rating is the MPA/age rating, e.g., PG, PG-13, R.\n\nIn the Oscar Awards Won column, list all of the qualifying Oscar categories above that the movie won, separated by commas. Do not include other awards.\nSort the list based on the number of qualifying Oscar awards won, in descending order.\nHighlight the highest IMDb score in the list with green fill + bold text.\nHighlight the lowest IMDb score with red fill + bold text.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns, at least 5 unique movies with no blank cells, and release years are spread at least 8 years apart.", "steps": [{"name": null, "description": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "text": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "text": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "text": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "text": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No cell in any of the required columns is left blank. (1 pt)", "text": "No cell in any of the required columns is left blank. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Accuracy", "description": "Each movie's genre, Oscar wins, IMDb score, and MPA rating are verified against authoritative sources. Each step is scored proportionally per movie: round((movies passing / total movies) * 10).", "steps": [{"name": null, "description": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "text": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Crime, Thriller). Word-boundary match: compound forms like \"Crime Drama\" pass for \"Crime\"; unrelated tags do not. (10 pt, proportional)", "text": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Crime, Thriller). Word-boundary match: compound forms like \"Crime Drama\" pass for \"Crime\"; unrelated tags do not. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each movie's release year must be between 1970 and 2010. (10 pt, proportional)", "text": "Each movie's release year must be between 1970 and 2010. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "text": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "text": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "text": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Every pair of release years in the sheet differs by at least 8 years. (10 pt, all or nothing)", "text": "Every pair of release years in the sheet differs by at least 8 years. (10 pt, all or nothing)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sorting and Conditional Formatting", "description": "The list is sorted by the number of qualifying Oscar awards won and the highest/lowest IMDb scores are visually highlighted.", "steps": [{"name": null, "description": "The movie list is sorted by the number of qualifying Oscar awards won in descending order. Ties may appear in any relative order. (1 pt)", "text": "The movie list is sorted by the number of qualifying Oscar awards won in descending order. Ties may appear in any relative order. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "text": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "text": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 15, "total_points": 78, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_55_movie_recommendation.5", "short_id": "55e", "family": "sheets_55_Movie_Recommendation", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Media Recommendation", "prompt": "I want to watch a movie tonight. I prefer the following genres: Drama, Comedy, Crime. Find the genre based on the IMDb listing of the movies.\nI want to watch award-winning movies, released between 1985 and 2005, and I prefer movies that are highly rated and well-reviewed (IMDb ≥ 6.5). Look for movies that have won at least one Oscar award for Best Actor, Best Actress, Best Adapted Screenplay, or Best Cinematography.\nPlease make a Google Sheet with at least 5 movie recommendations that won at least one of the mentioned awards. Each movie's primary genre on IMDb — the first standard genre IMDb lists for the movie — must be one of the above listed genres. Use IMDb's standard genre list, not UI tags. At least 3 different primary genres must appear across the sheet. The total duration of all movies must be between 600 and 800 minutes (inclusive).\nEach row should contain information about one movie. Use these column header names, in this order:\nMovie Title\nGenre\nMPA/Age Rating\nIMDb Score\nRelease Year\nDuration\nOscar Awards Won\n\nMPA/Age Rating is the MPA/age rating, e.g., PG, PG-13, R.\n\nIn the Oscar Awards Won column, list all of the qualifying Oscar categories above that the movie won, separated by commas. Do not include other awards.\nSort the list based on Duration in descending order.\nHighlight the highest IMDb score in the list with green fill + bold text.\nHighlight the lowest IMDb score with red fill + bold text.", "checkpoints": [{"n": 1, "title": "Spreadsheet Structure", "description": "The spreadsheet contains all required columns, at least 5 unique movies with no blank cells, includes at least 3 distinct primary genres, and the total duration falls within the required window.", "steps": [{"name": null, "description": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "text": "The sheet has all required labeled columns: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, and Oscar Awards Won. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "text": "The columns are in this exact order: Movie Title, Genre, MPA/Age Rating, IMDb Score, Release Year, Duration, Oscar Awards Won (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "text": "The sheet includes at least 5 movies (data rows, excluding the header). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "text": "Each row corresponds to one unique movie (no duplicate movie titles). (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No cell in any of the required columns is left blank. (1 pt)", "text": "No cell in any of the required columns is left blank. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Accuracy", "description": "Each movie's genre, Oscar wins, IMDb score, and MPA rating are verified against authoritative sources. Each step is scored proportionally per movie: round((movies passing / total movies) * 10).", "steps": [{"name": null, "description": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "text": "Each recommended movie's IMDb page (e.g., imdb.com/title/tt#####) appears in the browsing history. (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Comedy, Crime). Word-boundary match: compound forms like \"Crime Drama\" pass for \"Crime\"; unrelated tags do not. (10 pt, proportional)", "text": "Each recommended movie's primary genre on IMDb (the first standard genre IMDb lists for the movie, not UI tags) is one of the preferred genres (Drama, Comedy, Crime). Word-boundary match: compound forms like \"Crime Drama\" pass for \"Crime\"; unrelated tags do not. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "At least 3 distinct primary genres from the specified categories (per IMDb's standard genre list, first listed) appear among the recommended movies. (10 pt, all or nothing)", "text": "At least 3 distinct primary genres from the specified categories (per IMDb's standard genre list, first listed) appear among the recommended movies. (10 pt, all or nothing)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each movie's release year must be between 1985 and 2005. (10 pt, proportional)", "text": "Each movie's release year must be between 1985 and 2005. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "text": "The Oscar Awards Won cell lists exactly the qualifying Oscars (from {Best Actor, Best Actress, Best Adapted Screenplay, Best Cinematography}) that the movie actually won, comma-separated, with no extras and no omissions. The movie must have won at least one. Verified against the IMDb awards page. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "text": "Each IMDb Score in the sheet matches the movie's actual IMDb rating (within +/- 0.1 tolerance) and is >= 6.5. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "text": "Each MPA/Age Rating (e.g., PG, PG-13, R) is correct per the movie's IMDb listing or official distributor listing. (10 pt, proportional)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The sum of all movie durations is between 600 and 800 minutes, inclusive. (10 pt, all or nothing)", "text": "The sum of all movie durations is between 600 and 800 minutes, inclusive. (10 pt, all or nothing)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sorting and Conditional Formatting", "description": "The list is sorted by duration and the highest/lowest IMDb scores are visually highlighted.", "steps": [{"name": null, "description": "The movie list is sorted by Duration in descending order. (1 pt)", "text": "The movie list is sorted by Duration in descending order. (1 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "text": "The cell(s) with the highest IMDb Score have green background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "text": "The cell(s) with the lowest IMDb Score have red background fill and bold text. (1 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 3, "n_steps": 16, "total_points": 88, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_55_Movie_Recommendation/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_6_stock_tracker.1", "short_id": "6a", "family": "sheets_6_investmenttracker", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Investment Tracker", "prompt": "Make a stock tracker google sheet for the top 10 highest market cap stocks in tech from the end of Q2 2023. Insert a blank Google Sheets Table and then populate it with the data. Add the name and ticker of each of these to my watchlist and compare their past price to the current price for each stock to determine how much I would have gained or lost on each stock. Finally, put that I currently own 100 shares of each of these stocks and make a bar chart showing each of my stocks and their total value as a percentage of the overall portfolio.", "checkpoints": [{"n": 1, "title": "There is a spreadsheet with the requested information.", "description": "There is a spreadsheet with the requested information.", "steps": [{"name": null, "description": "There is a column/row with the name of each stock.", "text": "There is a column/row with the name of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the ticker symbol of each stock.", "text": "There is a column/row with the ticker symbol of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the current price of each stock.", "text": "There is a column/row with the current price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the past price of each stock.", "text": "There is a column/row with the past price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the gain/loss of each stock.", "text": "There is a column/row with the gain/loss of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row accounting for the number of shares owned for each stock.", "text": "There is a column/row accounting for the number of shares owned for each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the total value of each stock.", "text": "There is a column/row with the total value of each stock.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The stocks and information are correct and match the task requirements.", "description": "The stocks and information are correct and match the task requirements.", "steps": [{"name": null, "description": "Each of the 10 stocks matches a top 10 highest market cap stock in tech from the end of Q2 2023 (1 point per correct stock).", "text": "Each of the 10 stocks matches a top 10 highest market cap stock in tech from the end of Q2 2023 (1 point per correct stock).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The past price of each stock is correct.", "text": "The past price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The current price of each stock is correct.", "text": "The current price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The gain/loss of each stock is correct.", "text": "The gain/loss of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "A website containing the information was visited.", "description": "A website containing the information was visited.", "steps": [{"name": null, "description": "The website trace contains a URL that has information about the top 10 highest market cap stocks in tech from the end of Q2 2023.", "text": "The website trace contains a URL that has information about the top 10 highest market cap stocks in tech from the end of Q2 2023.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "The website trace contains a URL that has information about the past price of each stock.", "text": "The website trace contains a URL that has information about the past price of each stock.", "category": "web_visit_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "A bar chart showing each of the stocks and their total value as a percentage of the overall portfolio was created.", "description": "A bar chart showing each of the stocks and their total value as a percentage of the overall portfolio was created.", "steps": [{"name": null, "description": "A bar chart was created in the spreadsheet.", "text": "A bar chart was created in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The bar chart contains each of the stocks.", "text": "The bar chart contains each of the stocks.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bar chart correctly shows the total value of each stock as a percentage of the overall portfolio.", "text": "The bar chart correctly shows the total value of each stock as a percentage of the overall portfolio.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 25, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_6_stock_tracker.2", "short_id": "6b", "family": "sheets_6_investmenttracker", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Investment Tracker", "prompt": "Create a Google Sheet tracking the top 5 highest market cap companies in the consumer goods sector as of January 5, 2023. Insert a blank Google Sheets Table and then populate it with the data. List the name and ticker for each. Compare their closing price on that date to the current price to calculate the potential profit or loss per share. Assume I own 50 shares of each company and generate a pie chart displaying the current value of each holding as a percentage of the total portfolio.", "checkpoints": [{"n": 1, "title": "There is a spreadsheet with the requested information.", "description": "There is a spreadsheet with the requested information.", "steps": [{"name": null, "description": "There is a column/row with the name of each stock.", "text": "There is a column/row with the name of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the ticker symbol of each stock.", "text": "There is a column/row with the ticker symbol of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the current price of each stock.", "text": "There is a column/row with the current price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the past price of each stock.", "text": "There is a column/row with the past price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the gain/loss of each stock.", "text": "There is a column/row with the gain/loss of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row accounting for the number of shares owned for each stock.", "text": "There is a column/row accounting for the number of shares owned for each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the total value of each stock.", "text": "There is a column/row with the total value of each stock.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The stocks and information are correct and match the task requirements.", "description": "The stocks and information are correct and match the task requirements.", "steps": [{"name": null, "description": "Each of the 5 stocks matches a top 5 highest market cap company in the consumer goods sector as of January 5, 2023 (1 point per correct stock).", "text": "Each of the 5 stocks matches a top 5 highest market cap company in the consumer goods sector as of January 5, 2023 (1 point per correct stock).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The past price of each stock is correct.", "text": "The past price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The current price of each stock is correct.", "text": "The current price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The gain/loss of each stock is correct.", "text": "The gain/loss of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "A website containing the information was visited.", "description": "A website containing the information was visited.", "steps": [{"name": null, "description": "The website trace contains a URL that has information about the top 5 highest market cap companies in the consumer goods sector as of January 5, 2023.", "text": "The website trace contains a URL that has information about the top 5 highest market cap companies in the consumer goods sector as of January 5, 2023.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "The website trace contains a URL that has information about the past price of each stock.", "text": "The website trace contains a URL that has information about the past price of each stock.", "category": "web_visit_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "A pie chart showing each of the stocks and their current value as a percentage of the total portfolio was created.", "description": "A pie chart showing each of the stocks and their current value as a percentage of the total portfolio was created.", "steps": [{"name": null, "description": "A pie chart was created in the spreadsheet.", "text": "A pie chart was created in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The pie chart contains each of the stocks.", "text": "The pie chart contains each of the stocks.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The pie chart correctly shows the current value of each stock as a percentage of the overall portfolio.", "text": "The pie chart correctly shows the current value of each stock as a percentage of the overall portfolio.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 20, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_6_stock_tracker.3", "short_id": "6c", "family": "sheets_6_investmenttracker", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Investment Tracker", "prompt": "Make a stock tracker Google Sheet for the top 8 highest market cap banks recorded at the end of Q1 2024. Insert a blank Google Sheets Table and then populate it with the data. Add their names and tickers to the sheet. Determine the gain or loss by comparing the price at the end of Q1 2024 with today's price. Finally, calculate the total value assuming I hold 1,000 shares of each bank, and insert a bar chart showing the specific dollar value of each holding in the portfolio.", "checkpoints": [{"n": 1, "title": "There is a spreadsheet with the requested information.", "description": "There is a spreadsheet with the requested information.", "steps": [{"name": null, "description": "There is a column/row with the name of each stock.", "text": "There is a column/row with the name of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the ticker symbol of each stock.", "text": "There is a column/row with the ticker symbol of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the current price of each stock.", "text": "There is a column/row with the current price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the past price of each stock.", "text": "There is a column/row with the past price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the gain/loss of each stock.", "text": "There is a column/row with the gain/loss of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row accounting for the number of shares owned for each stock.", "text": "There is a column/row accounting for the number of shares owned for each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the total value of each stock.", "text": "There is a column/row with the total value of each stock.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The stocks and information are correct and match the task requirements.", "description": "The stocks and information are correct and match the task requirements.", "steps": [{"name": null, "description": "Each of the 8 stocks matches a top 8 highest market cap bank from the end of Q1 2024 (1 point per correct stock).", "text": "Each of the 8 stocks matches a top 8 highest market cap bank from the end of Q1 2024 (1 point per correct stock).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The past price of each stock is correct.", "text": "The past price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The current price of each stock is correct.", "text": "The current price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The gain/loss of each stock is correct.", "text": "The gain/loss of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "A website containing the information was visited.", "description": "A website containing the information was visited.", "steps": [{"name": null, "description": "The website trace contains a URL that has information about the top 8 highest market cap banks recorded at the end of Q1 2024.", "text": "The website trace contains a URL that has information about the top 8 highest market cap banks recorded at the end of Q1 2024.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "The website trace contains a URL that has information about the past price of each stock.", "text": "The website trace contains a URL that has information about the past price of each stock.", "category": "web_visit_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "A bar chart showing the specific dollar value of each holding in the portfolio was created.", "description": "A bar chart showing the specific dollar value of each holding in the portfolio was created.", "steps": [{"name": null, "description": "A bar chart was created in the spreadsheet.", "text": "A bar chart was created in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The bar chart contains each of the stocks.", "text": "The bar chart contains each of the stocks.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The bar chart correctly shows the dollar value of each holding in the portfolio.", "text": "The bar chart correctly shows the dollar value of each holding in the portfolio.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 23, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_6_stock_tracker.4", "short_id": "6d", "family": "sheets_6_investmenttracker", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Investment Tracker", "prompt": "Create a Google Sheet for the top 7 semiconductor companies by market cap as of December 31, 2025. Insert a blank Google Sheets Table and then populate it with the data. Include the company name and ticker symbol. Calculate the difference between the closing price on that date and the current market price to show my theoretical return. I currently own 25 shares of each stock; create a line chart that visualizes the current total value of each position relative to the others.", "checkpoints": [{"n": 1, "title": "There is a spreadsheet with the requested information.", "description": "There is a spreadsheet with the requested information.", "steps": [{"name": null, "description": "There is a column/row with the name of each stock.", "text": "There is a column/row with the name of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the ticker symbol of each stock.", "text": "There is a column/row with the ticker symbol of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the current price of each stock.", "text": "There is a column/row with the current price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the past price of each stock.", "text": "There is a column/row with the past price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the gain/loss of each stock.", "text": "There is a column/row with the gain/loss of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row accounting for the number of shares owned for each stock.", "text": "There is a column/row accounting for the number of shares owned for each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the total value of each stock.", "text": "There is a column/row with the total value of each stock.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The stocks and information are correct and match the task requirements.", "description": "The stocks and information are correct and match the task requirements.", "steps": [{"name": null, "description": "Each of the 7 stocks matches a top 7 highest market cap semiconductor company as of December 31, 2025 (1 point per correct stock).", "text": "Each of the 7 stocks matches a top 7 highest market cap semiconductor company as of December 31, 2025 (1 point per correct stock).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The past price of each stock is correct.", "text": "The past price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The current price of each stock is correct.", "text": "The current price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The gain/loss of each stock is correct.", "text": "The gain/loss of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "A website containing the information was visited.", "description": "A website containing the information was visited.", "steps": [{"name": null, "description": "The website trace contains a URL that has information about the top 7 semiconductor companies by market cap as of December 31, 2025.", "text": "The website trace contains a URL that has information about the top 7 semiconductor companies by market cap as of December 31, 2025.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "The website trace contains a URL that has information about the past price of each stock.", "text": "The website trace contains a URL that has information about the past price of each stock.", "category": "web_visit_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "A line chart showing the current total value of each position relative to the others was created.", "description": "A line chart showing the current total value of each position relative to the others was created.", "steps": [{"name": null, "description": "A line chart was created in the spreadsheet.", "text": "A line chart was created in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The line chart contains each of the stocks.", "text": "The line chart contains each of the stocks.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The line chart correctly shows the current total value of each position.", "text": "The line chart correctly shows the current total value of each position.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 22, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_6_stock_tracker.5", "short_id": "6e", "family": "sheets_6_investmenttracker", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Investment Tracker", "prompt": "Build a stock tracking Google Sheet for the top 6 pharmaceutical companies by market cap as of 05/19/2025. Insert a blank Google Sheets Table and then populate it with the data. Log the name and ticker for each. Compare the past price from that date to the current price to determine the gain or loss per stock. Assume I possess 250 shares of each and create a pie chart illustrating each stock's current value as a percentage of the total portfolio sum.", "checkpoints": [{"n": 1, "title": "There is a spreadsheet with the requested information.", "description": "There is a spreadsheet with the requested information.", "steps": [{"name": null, "description": "There is a column/row with the name of each stock.", "text": "There is a column/row with the name of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the ticker symbol of each stock.", "text": "There is a column/row with the ticker symbol of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the current price of each stock.", "text": "There is a column/row with the current price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the past price of each stock.", "text": "There is a column/row with the past price of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the gain/loss of each stock.", "text": "There is a column/row with the gain/loss of each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row accounting for the number of shares owned for each stock.", "text": "There is a column/row accounting for the number of shares owned for each stock.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "There is a column/row with the total value of each stock.", "text": "There is a column/row with the total value of each stock.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The stocks and information are correct and match the task requirements.", "description": "The stocks and information are correct and match the task requirements.", "steps": [{"name": null, "description": "Each of the 6 stocks matches a top 6 highest market cap pharmaceutical company as of 05-19-2025 (1 point per correct stock).", "text": "Each of the 6 stocks matches a top 6 highest market cap pharmaceutical company as of 05-19-2025 (1 point per correct stock).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The past price of each stock is correct.", "text": "The past price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The current price of each stock is correct.", "text": "The current price of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The gain/loss of each stock is correct.", "text": "The gain/loss of each stock is correct.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "A website containing the information was visited.", "description": "A website containing the information was visited.", "steps": [{"name": null, "description": "The website trace contains a URL that has information about the top 6 pharmaceutical companies by market cap as of 05/19/2025.", "text": "The website trace contains a URL that has information about the top 6 pharmaceutical companies by market cap as of 05/19/2025.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "The website trace contains a URL that has information about the past price of each stock.", "text": "The website trace contains a URL that has information about the past price of each stock.", "category": "web_visit_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "A pie chart showing each stock's current value as a percentage of the total portfolio sum was created.", "description": "A pie chart showing each stock's current value as a percentage of the total portfolio sum was created.", "steps": [{"name": null, "description": "A pie chart was created in the spreadsheet.", "text": "A pie chart was created in the spreadsheet.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The pie chart contains each of the stocks.", "text": "The pie chart contains each of the stocks.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The pie chart correctly shows the current value of each stock as a percentage of the overall portfolio.", "text": "The pie chart correctly shows the current value of each stock as a percentage of the overall portfolio.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 21, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_6_investmenttracker/instance_5/checkpoints.md"}
{"task_id": "knows.sheets_7_running_analysis.1", "short_id": "7a", "family": "sheets_7_running_analysis", "instance": "instance_1", "instance_num": 1, "artifact": "Sheets", "group": "Personal", "domain": "Fitness Analytics", "prompt": "Gather my running data from the provided Drive link: https://drive.google.com/drive/folders/1YM6wCpN1c2NjgWs7De4o67obg2eSCdZ5?usp=sharing and put relevant data into a table in a Google Sheet. Create 2 plots using this data, one that shows my average running speed in (min/mile) over time, and one that shows the cumulative distance I've ran in miles over time. The running speed plot should have all runs as circular points and the distance plot should be a tracking the increasing distance over time. Include these plots in the same sheet tab as the data table. Find online and add in the average male age 25 running speed for a 5k as a baseline dotted line in the average running speed plot. Compare this to the average running speed for Eliud Kipchoge in his top 3 races in the Marathon. Make sure everything in the google sheet is fully unobscured and put the sources for the average running times you retrieve below the relevant plot.", "checkpoints": [{"n": 1, "title": "Data Table Structure", "description": "The spreadsheet contains a properly structured table with required columns for running data.", "steps": [{"name": null, "description": "Date/time column exists with a header containing \"date\" or \"activity date\".", "text": "Date/time column exists with a header containing \"date\" or \"activity date\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "text": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "text": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All table content is fully visible (no text overflow/truncation).", "text": "All table content is fully visible (no text overflow/truncation).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Table Content Accuracy", "description": "The data in the table matches the source running data for all 109 Run activities. Each step is scored out of 10 points using `floor(matches/109 * 10)`. The boolean PASS state requires every row to match exactly.", "steps": [{"name": null, "description": "Run activity rows have an exact date match to gold data (10 pts, proportional credit).", "text": "Run activity rows have an exact date match to gold data (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Run activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "text": "Run activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Run activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "text": "Run activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Speed Over Time Plot", "description": "A scatter plot showing average running speed (min/mile) over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates speed (min/mile or similar).", "text": "Y-axis label indicates speed (min/mile or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates speed over time.", "text": "Chart title indicates speed over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart main data series comes from the average speed column.", "text": "Chart main data series comes from the average speed column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speed values are present as circular points in the chart.", "text": "Speed values are present as circular points in the chart.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Male 5K baseline is properly displayed (labeled in legend + dotted/dashed style).", "text": "Male 5K baseline is properly displayed (labeled in legend + dotted/dashed style).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Male 5K baseline data is constant and within expected range (7.1-10 min/mile).", "text": "Male 5K baseline data is constant and within expected range (7.1-10 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Kipchoge baseline is properly displayed (labeled in legend).", "text": "Kipchoge baseline is properly displayed (labeled in legend).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Kipchoge baseline data is constant and within expected range (4.5-4.8 min/mile).", "text": "Kipchoge baseline data is constant and within expected range (4.5-4.8 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Both baselines are visually distinguishable from the main data.", "text": "Both baselines are visually distinguishable from the main data.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Source URLs are valid and accessible below the speed chart.", "text": "Source URLs are valid and accessible below the speed chart.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Cumulative Distance Plot", "description": "A chart showing cumulative distance ran in miles over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates cumulative distance (miles or similar).", "text": "Y-axis label indicates cumulative distance (miles or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates cumulative distance over time.", "text": "Chart title indicates cumulative distance over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Data shows cumulative/running total.", "text": "Data shows cumulative/running total.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Cumulative running values are present as a line plot.", "text": "Cumulative running values are present as a line plot.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather data. URL relevance is determined first by keyword matching against the URL string, with an LLM-as-judge backup that fetches and inspects page content when the keyword filter returns no candidates. Content validity is determined first by structured LLM pace extraction + numeric comparison against the sheet baseline (≤5% tolerance), with an LLM-as-judge backup that asks the model whether the page supports a baseline within ~5% of the sheet value when extraction fails for every URL.", "steps": [{"name": null, "description": "A source URL for male 5K running speed was visited (keyword match, then LLM judge backup).", "text": "A source URL for male 5K running speed was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "A source URL for Eliud Kipchoge marathon data was visited (keyword match, then LLM judge backup).", "text": "A source URL for Eliud Kipchoge marathon data was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Male 5K source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Male 5K source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "Kipchoge source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Kipchoge source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}]}], "n_checkpoints": 5, "n_steps": 31, "total_points": 58, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1YM6wCpN1c2NjgWs7De4o67obg2eSCdZ5?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_1/checkpoints.md"}
{"task_id": "knows.sheets_7_running_analysis.2", "short_id": "7b", "family": "sheets_7_running_analysis", "instance": "instance_2", "instance_num": 2, "artifact": "Sheets", "group": "Personal", "domain": "Fitness Analytics", "prompt": "Gather my running data from the provided Drive link: https://drive.google.com/drive/folders/1XS9HknQMyosKt_yiO8Xiz8R6LkmTsQTm and put relevant data into a table in a Google Sheet. Create 2 plots using this data, one that shows my average running speed in (min/mile) over time, and one that shows the cumulative distance I've ran in miles over time. The running speed plot should have all runs as circular points and the distance plot should be a tracking the increasing distance over time. Include these plots in the same sheet tab as the data table. Find online and add in the average intermediate male 30 running speed for a half-marathon as a baseline dotted line in the average running speed plot. Compare this to the average running speed for Jacob Kiplimo in his top 5 races in the Half Marathon. Make sure everything in the google sheet is fully unobscured and put the sources for the average running times you retrieve below the relevant plot.", "checkpoints": [{"n": 1, "title": "Data Table Structure", "description": "The spreadsheet contains a properly structured table with required columns for running data.", "steps": [{"name": null, "description": "Date/time column exists with a header containing \"date\" or \"activity date\".", "text": "Date/time column exists with a header containing \"date\" or \"activity date\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "text": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "text": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All table content is fully visible (no text overflow/truncation).", "text": "All table content is fully visible (no text overflow/truncation).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Table Content Accuracy", "description": "The data in the table matches the source running data for all 169 Run activities. Each step is scored out of 10 points using `floor(matches/169 * 10)`. The boolean PASS state requires every row to match exactly.", "steps": [{"name": null, "description": "Run activity rows have an exact date match to gold data (10 pts, proportional credit).", "text": "Run activity rows have an exact date match to gold data (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Run activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "text": "Run activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Run activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "text": "Run activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Speed Over Time Plot", "description": "A scatter plot showing average running speed (min/mile) over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates speed (min/mile or similar).", "text": "Y-axis label indicates speed (min/mile or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates speed over time.", "text": "Chart title indicates speed over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart main data series comes from the average speed column.", "text": "Chart main data series comes from the average speed column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speed values are present as circular points in the chart.", "text": "Speed values are present as circular points in the chart.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Intermediate Male 30 Half-Marathon baseline is properly displayed (labeled in legend + dotted/dashed style).", "text": "Intermediate Male 30 Half-Marathon baseline is properly displayed (labeled in legend + dotted/dashed style).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Intermediate Male 30 Half-Marathon baseline data is constant and within expected range (7.0–9.0 min/mile).", "text": "Intermediate Male 30 Half-Marathon baseline data is constant and within expected range (7.0–9.0 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Jacob Kiplimo Half Marathon baseline is properly displayed (labeled in legend).", "text": "Jacob Kiplimo Half Marathon baseline is properly displayed (labeled in legend).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Jacob Kiplimo Half Marathon baseline data is constant and within expected range (4.15–4.45 min/mile).", "text": "Jacob Kiplimo Half Marathon baseline data is constant and within expected range (4.15–4.45 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Both baselines are visually distinguishable from the main data.", "text": "Both baselines are visually distinguishable from the main data.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Source URLs are valid and accessible below the speed chart.", "text": "Source URLs are valid and accessible below the speed chart.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Cumulative Distance Plot", "description": "A chart showing cumulative distance ran in miles over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates cumulative distance (miles or similar).", "text": "Y-axis label indicates cumulative distance (miles or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates cumulative distance over time.", "text": "Chart title indicates cumulative distance over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Data shows cumulative/running total.", "text": "Data shows cumulative/running total.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Cumulative running values are present as a line plot.", "text": "Cumulative running values are present as a line plot.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather data. URL relevance is determined first by keyword matching against the URL string, with an LLM-as-judge backup that fetches and inspects page content when the keyword filter returns no candidates. Content validity is determined first by structured LLM pace extraction + numeric comparison against the sheet baseline (≤5% tolerance), with an LLM-as-judge backup that asks the model whether the page supports a baseline within ~5% of the sheet value when extraction fails for every URL.", "steps": [{"name": null, "description": "A source URL for intermediate male 30 half-marathon running speed was visited (keyword match, then LLM judge backup).", "text": "A source URL for intermediate male 30 half-marathon running speed was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "A source URL for Jacob Kiplimo half marathon data was visited (keyword match, then LLM judge backup).", "text": "A source URL for Jacob Kiplimo half marathon data was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Intermediate Male 30 Half-Marathon source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Intermediate Male 30 Half-Marathon source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "Jacob Kiplimo source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Jacob Kiplimo source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}]}], "n_checkpoints": 5, "n_steps": 31, "total_points": 58, "requires_browsing_history": true, "source_assets": ["https://drive.google.com/drive/folders/1XS9HknQMyosKt_yiO8Xiz8R6LkmTsQTm"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_2/checkpoints.md"}
{"task_id": "knows.sheets_7_running_analysis.3", "short_id": "7c", "family": "sheets_7_running_analysis", "instance": "instance_3", "instance_num": 3, "artifact": "Sheets", "group": "Personal", "domain": "Fitness Analytics", "prompt": "Gather my Nordic skiing data from this Drive link: https://docs.google.com/spreadsheets/d/1XjcWMfSKnxuNynDnwHut-JyKTbWKhT8CFPMflgRi6O8/edit?usp=sharing and put the relevant data into a table in a Google Sheet. Create two plots from the data:\n\nAverage skiing pace (min/mile) per run over time — each run as a circular point. Add a dotted baseline line showing Johannes Høsflot Klæbo's average pace across the 20 km skiathlon and the sprint classic in the 2026 Olympics.\n\nCumulative distance skied (miles) over time as a tracking line that climbs as runs accumulate.\n\nInclude both plots in the same sheet tab as the data table, fully unobscured. Below the pace plot, cite the source(s) for Klæbo's race times — race name, year, technique, and finishing time.", "checkpoints": [{"n": 1, "title": "Data Table Structure", "description": "The spreadsheet contains a properly structured table with required columns for Nordic skiing data.", "steps": [{"name": null, "description": "Date/time column exists with a header containing \"date\" or \"activity date\".", "text": "Date/time column exists with a header containing \"date\" or \"activity date\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "text": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "text": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All table content is fully visible (no text overflow/truncation).", "text": "All table content is fully visible (no text overflow/truncation).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Table Content Accuracy", "description": "The data in the table matches the source skiing data for all 34 Nordic Ski activities. Each step is scored out of 10 points using `floor(matches/34 * 10)`. The boolean PASS state requires every row to match exactly.", "steps": [{"name": null, "description": "Nordic Ski activity rows have an exact date match to gold data (10 pts, proportional credit).", "text": "Nordic Ski activity rows have an exact date match to gold data (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Nordic Ski activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "text": "Nordic Ski activity rows have an exact distance match to gold data, converted to miles (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Nordic Ski activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "text": "Nordic Ski activity rows have an exact average speed match to gold data, converted to min/mile (10 pts, proportional credit).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Pace Over Time Plot", "description": "A scatter plot showing average skiing pace (min/mile) over time, with a single dotted Klæbo baseline.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates pace (min/mile or similar).", "text": "Y-axis label indicates pace (min/mile or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates pace over time.", "text": "Chart title indicates pace over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart main data series comes from the average speed column.", "text": "Chart main data series comes from the average speed column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Pace values are present as circular points in the chart.", "text": "Pace values are present as circular points in the chart.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Klæbo baseline is properly displayed (labeled in legend + dotted/dashed style).", "text": "Klæbo baseline is properly displayed (labeled in legend + dotted/dashed style).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Klæbo baseline data is constant and within expected range (3.6–3.9 min/mile).", "text": "Klæbo baseline data is constant and within expected range (3.6–3.9 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Klæbo baseline is visually distinguishable from the main data.", "text": "Klæbo baseline is visually distinguishable from the main data.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Source URL(s) are valid and accessible below the pace chart.", "text": "Source URL(s) are valid and accessible below the pace chart.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Cumulative Distance Plot", "description": "A chart showing cumulative distance skied in miles over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates cumulative distance (miles or similar).", "text": "Y-axis label indicates cumulative distance (miles or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates cumulative distance over time.", "text": "Chart title indicates cumulative distance over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Data shows cumulative/running total.", "text": "Data shows cumulative/running total.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Cumulative distance values are present as a line plot.", "text": "Cumulative distance values are present as a line plot.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is on the same sheet tab as the data table.", "text": "Chart is on the same sheet tab as the data table.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Website Visit Validation", "description": "The agent visited the required website to gather Klæbo's race data. URL relevance is determined first by keyword matching against the URL string, with an LLM-as-judge backup that fetches and inspects page content when the keyword filter returns no candidates. Content validity is determined first by structured LLM pace extraction + numeric comparison against the sheet baseline (≤5% tolerance), with an LLM-as-judge backup that asks the model whether the page supports a baseline within ~5% of the sheet value when extraction fails for every URL.", "steps": [{"name": null, "description": "A source URL for Klæbo's race data was visited (keyword match, then LLM judge backup).", "text": "A source URL for Klæbo's race data was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Klæbo source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Klæbo source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}]}], "n_checkpoints": 5, "n_steps": 27, "total_points": 54, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/spreadsheets/d/1XjcWMfSKnxuNynDnwHut-JyKTbWKhT8CFPMflgRi6O8/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_3/checkpoints.md"}
{"task_id": "knows.sheets_7_running_analysis.4", "short_id": "7d", "family": "sheets_7_running_analysis", "instance": "instance_4", "instance_num": 4, "artifact": "Sheets", "group": "Personal", "domain": "Fitness Analytics", "prompt": "Gather my running  and walking data from the provided Drive link: https://docs.google.com/spreadsheets/d/1I7db2Xx1nWauYqixEoxmMUWyt0ojME8LEJDbbbEsH_0/edit?usp=sharing and put relevant data into a table in a Google Sheet. Create 3 plots using this data, one that shows my daily total miles,one that shows my average running speed in (min/mile) over time, and one that shows the cumulative distance I've ran in miles over time. The daily total miles plot should have all walk and run data as a bar plot and the distance plot should be a tracking the increasing distance over time. Include these plots in the same Google Sheet. Find online and add in the average female 25 average daily miles trekked as a baseline dotted line in the daily total miles plot. Then also find online and add in the average female 25 running speed for a 5k as a baseline dotted line in the average running speed plot. Compare this to the average running speed for Beatrice Chebet in the 5k. Make sure everything in the google sheet is fully unobscured and put the sources for the average running times you retrieve below the relevant plot.", "checkpoints": [{"n": 1, "title": "Data Table Structure", "description": "The spreadsheet contains a properly structured table with required columns for running and walking data.", "steps": [{"name": null, "description": "Date/time column exists with a header containing \"date\" or \"activity date\".", "text": "Date/time column exists with a header containing \"date\" or \"activity date\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "text": "Distance column exists with a header containing \"distance\", \"miles\", or an equivalent miles-unit token (e.g., \"(miles)\", \"(mi)\").", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "text": "Average Speed column exists with a header containing \"speed\", \"pace\", or \"min/mile\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All table content is fully visible (no text overflow/truncation).", "text": "All table content is fully visible (no text overflow/truncation).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Table Content Accuracy", "description": "The data in the table matches the source data for all 452 Run and Walk activities. Each step is scored proportionally to the number of rows that match: a step's awarded points equal `floor(matches/452 * 10)` out of 10 (e.g. 440/452 -> 9/10 pts). The boolean PASS state still requires every row to match exactly.", "steps": [{"name": null, "description": "Activity rows have an exact date match to gold data (proportional credit, 10 pts max).", "text": "Activity rows have an exact date match to gold data (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Activity rows have an exact distance match to gold data, converted to miles (proportional credit, 10 pts max).", "text": "Activity rows have an exact distance match to gold data, converted to miles (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Activity rows have an exact average speed match to gold data, converted to min/mile (proportional credit, 10 pts max).", "text": "Activity rows have an exact average speed match to gold data, converted to min/mile (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Daily Total Miles Bar Plot", "description": "A bar chart showing daily total miles with both walk and run data.", "steps": [{"name": null, "description": "X-axis label indicates date.", "text": "X-axis label indicates date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates miles or distance.", "text": "Y-axis label indicates miles or distance.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates daily total miles.", "text": "Chart title indicates daily total miles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is a BAR type.", "text": "Chart is a BAR type.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart has a non-constant data series representing daily mileage.", "text": "Chart has a non-constant data series representing daily mileage.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart data includes both walk and run activity data (data point count >= run count).", "text": "Chart data includes both walk and run activity data (data point count >= run count).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Female 25 daily miles baseline is properly displayed (labeled in legend).", "text": "Female 25 daily miles baseline is properly displayed (labeled in legend).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Female 25 daily miles baseline data is constant and within expected range (1.5-6.0 miles).", "text": "Female 25 daily miles baseline data is constant and within expected range (1.5-6.0 miles).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "Speed Over Time Plot", "description": "A plot showing average running speed (min/mile) over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates speed (min/mile or similar).", "text": "Y-axis label indicates speed (min/mile or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates speed over time.", "text": "Chart title indicates speed over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart main data series comes from the average speed column.", "text": "Chart main data series comes from the average speed column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Female 25 5K baseline is properly displayed (labeled in legend + dotted/dashed style).", "text": "Female 25 5K baseline is properly displayed (labeled in legend + dotted/dashed style).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Female 25 5K baseline data is constant and within expected range (7.5-11.5 min/mile).", "text": "Female 25 5K baseline data is constant and within expected range (7.5-11.5 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Beatrice Chebet baseline is properly displayed (labeled in legend).", "text": "Beatrice Chebet baseline is properly displayed (labeled in legend).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Beatrice Chebet baseline data is constant and within expected range (4.0-5.0 min/mile).", "text": "Beatrice Chebet baseline data is constant and within expected range (4.0-5.0 min/mile).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Both baselines are visually distinguishable from the main data.", "text": "Both baselines are visually distinguishable from the main data.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Source URLs are valid and accessible below the speed chart.", "text": "Source URLs are valid and accessible below the speed chart.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Cumulative Distance Plot", "description": "A chart showing cumulative distance ran in miles over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates cumulative distance (miles or similar).", "text": "Y-axis label indicates cumulative distance (miles or similar).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates cumulative distance over time.", "text": "Chart title indicates cumulative distance over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Data shows cumulative/running total.", "text": "Data shows cumulative/running total.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Cumulative running values are present as a line plot.", "text": "Cumulative running values are present as a line plot.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 6, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather data. URL relevance is determined first by keyword matching against the URL string, with an LLM-as-judge backup that fetches and inspects page content when the keyword filter returns no candidates. Content validity is determined first by structured LLM pace extraction + numeric comparison against the sheet baseline (<=5% tolerance), with an LLM-as-judge backup that asks the model whether the page supports a baseline within ~5% of the sheet value when extraction fails for every URL.", "steps": [{"name": null, "description": "A source URL for average female 25 daily miles trekked was visited (keyword match, then LLM judge backup).", "text": "A source URL for average female 25 daily miles trekked was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "A source URL for average female 25 5K running speed was visited (keyword match, then LLM judge backup).", "text": "A source URL for average female 25 5K running speed was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "A source URL for Beatrice Chebet 5K data was visited (keyword match, then LLM judge backup).", "text": "A source URL for Beatrice Chebet 5K data was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Female daily miles source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Female daily miles source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "Female 5K source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Female 5K source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "Beatrice Chebet source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "text": "Beatrice Chebet source URL content matches the sheet baseline within 5% (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}]}], "n_checkpoints": 6, "n_steps": 39, "total_points": 66, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/spreadsheets/d/1I7db2Xx1nWauYqixEoxmMUWyt0ojME8LEJDbbbEsH_0/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_4/checkpoints.md"}
{"task_id": "knows.sheets_7_running_analysis.5", "short_id": "7e", "family": "sheets_7_running_analysis", "instance": "instance_5", "instance_num": 5, "artifact": "Sheets", "group": "Personal", "domain": "Fitness Analytics", "prompt": "Gather my strength training data from the provided Drive link: https://docs.google.com/spreadsheets/d/1V8kYA_IOaLK-4CA-yp7Ne-ZvpIhEZZbsBgBaY_9sZrY/edit?gid=1825386740#gid=1825386740 and put relevant data into a table in a Google Sheet. Create 2 plots using this data, one that shows my individual workout sets over time, and one that shows the cumulative reps I've lifted over time. The individual workout sets plot should have all workouts as circular points and the cumulative sets  plot should be a tracking the increasing reps over time. Include these plots in the same Google Sheet. Find online and add in the average adult workout sets as a baseline dotted line in the individual workout sets plot. Compare this to the average sets for Jay Cutler per workout. Make sure everything in the google sheet is fully unobscured and put the sources for the average workout sets you retrieve below the relevant plot.", "checkpoints": [{"n": 1, "title": "Data Table Structure", "description": "The spreadsheet contains a properly structured table with required columns for strength training data.", "steps": [{"name": null, "description": "Date/time column exists with a header containing \"date\" or \"activity date\".", "text": "Date/time column exists with a header containing \"date\" or \"activity date\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Total Sets column exists with a header containing \"sets\" or \"total sets\".", "text": "Total Sets column exists with a header containing \"sets\" or \"total sets\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Total Reps column exists with a header containing \"reps\" or \"total reps\".", "text": "Total Reps column exists with a header containing \"reps\" or \"total reps\".", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All table content is fully visible (no text overflow/truncation).", "text": "All table content is fully visible (no text overflow/truncation).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Data Table Content Accuracy", "description": "The data in the table matches the source strength training data for all 219 Strength Training activities. Each step is scored proportionally: `floor(matches/219 * 10)` points out of 10 (e.g. 210/219 -> 9/10 pts). The boolean PASS state still requires every row to match exactly.", "steps": [{"name": null, "description": "Activity rows have an exact date match to gold data (proportional credit, 10 pts max).", "text": "Activity rows have an exact date match to gold data (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Activity rows have an exact Total Sets match to gold data (proportional credit, 10 pts max).", "text": "Activity rows have an exact Total Sets match to gold data (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Activity rows have an exact Total Reps match to gold data (proportional credit, 10 pts max).", "text": "Activity rows have an exact Total Reps match to gold data (proportional credit, 10 pts max).", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 3, "title": "Sets Over Time Plot", "description": "A chart showing individual workout sets over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates sets.", "text": "Y-axis label indicates sets.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates sets over time.", "text": "Chart title indicates sets over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart main data series comes from the Total Sets column.", "text": "Chart main data series comes from the Total Sets column.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Set values are present as circular points in the chart.", "text": "Set values are present as circular points in the chart.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Average adult sets baseline is properly displayed (labeled in legend + dotted/dashed style).", "text": "Average adult sets baseline is properly displayed (labeled in legend + dotted/dashed style).", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Average adult sets baseline data is constant and within expected range (15-22 sets).", "text": "Average adult sets baseline data is constant and within expected range (15-22 sets).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Jay Cutler baseline is properly displayed (labeled in legend).", "text": "Jay Cutler baseline is properly displayed (labeled in legend).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Jay Cutler baseline data is constant and within expected range (40-80 sets).", "text": "Jay Cutler baseline data is constant and within expected range (40-80 sets).", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Both baselines are visually distinguishable from the main data.", "text": "Both baselines are visually distinguishable from the main data.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Source URLs are valid and accessible below the sets chart.", "text": "Source URLs are valid and accessible below the sets chart.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Cumulative Reps Plot", "description": "A chart showing cumulative reps lifted over time.", "steps": [{"name": null, "description": "X-axis label indicates activity date.", "text": "X-axis label indicates activity date.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Y-axis label indicates cumulative reps.", "text": "Y-axis label indicates cumulative reps.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart title indicates cumulative reps over time.", "text": "Chart title indicates cumulative reps over time.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Chart is not placed over any other charts or tables.", "text": "Chart is not placed over any other charts or tables.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Data shows cumulative/running total.", "text": "Data shows cumulative/running total.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Cumulative reps values are present as a line plot.", "text": "Cumulative reps values are present as a line plot.", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 5, "title": "Website Visit Validation", "description": "The agent visited the required websites to gather data. URL relevance is determined first by keyword matching against the URL string, with an LLM-as-judge backup that fetches and inspects page content when the keyword filter returns no candidates. Content validity is determined first by structured LLM extraction + numeric comparison against the sheet baseline, with an LLM-as-judge backup that asks the model whether the page supports a baseline within the acceptable range when extraction fails for every URL.", "steps": [{"name": null, "description": "A source URL for average adult workout sets was visited (keyword match, then LLM judge backup).", "text": "A source URL for average adult workout sets was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "A source URL for Jay Cutler workout sets data was visited (keyword match, then LLM judge backup).", "text": "A source URL for Jay Cutler workout sets data was visited (keyword match, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Average adult sets source URL content matches the sheet baseline within acceptable range (pace extraction, then LLM judge backup).", "text": "Average adult sets source URL content matches the sheet baseline within acceptable range (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "Jay Cutler source URL content matches the sheet baseline within acceptable range (pace extraction, then LLM judge backup).", "text": "Jay Cutler source URL content matches the sheet baseline within acceptable range (pace extraction, then LLM judge backup).", "category": "web_visit_check", "llm_vlm_primary": true}]}], "n_checkpoints": 5, "n_steps": 29, "total_points": 56, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/spreadsheets/d/1V8kYA_IOaLK-4CA-yp7Ne-ZvpIhEZZbsBgBaY_9sZrY/edit?gid=1825386740#gid=1825386740"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/sheets_7_running_analysis/instance_5/checkpoints.md"}
{"task_id": "knows.slides_17_remove_images_add_placeholders.1", "short_id": "17a", "family": "slides_17_removeimagesaddplaceholders", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Cross", "domain": "Slide Editing", "prompt": "Take the presentation at https://docs.google.com/presentation/d/1oYvhqR6AtsYhpZSJwxnG5Zs6Xum1Gk5wNOVIJkSU4Fo/edit and help me replace all the images. Here is a copy of the presentation for you to edit: {{WORKING_COPY_URL}}. Identify all of the images in the document and save them all as image files and place them into {{IMAGES_FOLDER_URL}} for possible later use. Delete all the images, replacing each one with a description of the image in a text box at the same spot in big red text. Finally, for each of the images, find a similar image online and place it in the right spot. For each of these, make sure to put the URL for the image address below each new image to credit it.", "checkpoints": [{"n": 1, "title": "All original images from the presentation are saved to the specified Drive folder.", "description": "All original images from the presentation are saved to the specified Drive folder.", "steps": [{"name": null, "description": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "text": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "At each original image location, the image has been removed (10pt).", "text": "At each original image location, the image has been removed (10pt).", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "text": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "text": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "text": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "text": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Each replacement image has a URL credit beneath it (10pt).", "text": "Each replacement image has a URL credit beneath it (10pt).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "text": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "text": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The replacement image fully covers the text placeholder so no red text is visible - 80% coverage required (10pt).", "text": "The replacement image fully covers the text placeholder so no red text is visible - 80% coverage required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "description": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "steps": [{"name": null, "description": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "text": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "text": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 12, "total_points": 120, "requires_browsing_history": false, "source_assets": ["https://docs.google.com/presentation/d/1oYvhqR6AtsYhpZSJwxnG5Zs6Xum1Gk5wNOVIJkSU4Fo/edit"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_1/checkpoints.md"}
{"task_id": "knows.slides_17_remove_images_add_placeholders.2", "short_id": "17b", "family": "slides_17_removeimagesaddplaceholders", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Cross", "domain": "Slide Editing", "prompt": "Take the presentation at https://docs.google.com/presentation/d/12J_yNv6goL1q5Ctd5ai8Rysig-u3L-J6WZbx0Koib8M/edit and help me replace all the images. First, make a copy of the original presentation and put in into {{OUTPUT_FOLDER_URL}}. Then identify all of the images in the document and save them all as image files and place them into {{IMAGES_FOLDER_URL}} for possible later use. Delete all the images, replacing each one with a description of the image in a text box at the same spot in big red text. Finally, for each of the images, find a similar image online and place it in the right spot. For each of these, make sure to put the URL for the image address below each new image to credit it.", "checkpoints": [{"n": 1, "title": "All original images from the presentation are saved to the specified Drive folder.", "description": "All original images from the presentation are saved to the specified Drive folder.", "steps": [{"name": null, "description": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "text": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "At each original image location, the image has been removed (10pt).", "text": "At each original image location, the image has been removed (10pt).", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "text": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "text": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "text": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "text": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Each replacement image has a URL credit beneath it (10pt).", "text": "Each replacement image has a URL credit beneath it (10pt).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "text": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "text": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "text": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "description": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "steps": [{"name": null, "description": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "text": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "text": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 12, "total_points": 120, "requires_browsing_history": false, "source_assets": ["https://docs.google.com/presentation/d/12J_yNv6goL1q5Ctd5ai8Rysig-u3L-J6WZbx0Koib8M/edit"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_2/checkpoints.md"}
{"task_id": "knows.slides_17_remove_images_add_placeholders.3", "short_id": "17c", "family": "slides_17_removeimagesaddplaceholders", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Cross", "domain": "Slide Editing", "prompt": "Take the presentation at https://docs.google.com/presentation/d/1NAAZwHMwKjXt3A1BfZ4OJiUGWKaiv7Vx0c21sa1SW1E/edit and help me replace all the images. First, make a copy of the original presentation and put in into {{OUTPUT_FOLDER_URL}}. Then identify all of the images in the document and save them all as image files and place them into {{IMAGES_FOLDER_URL}} for possible later use. Delete all the images, replacing each one with a description of the image in a text box at the same spot in big red text. Finally, for each of the images, find a similar image online and place it in the right spot. For each of these, make sure to put the URL for the image address below each new image to credit it.", "checkpoints": [{"n": 1, "title": "All original images from the presentation are saved to the specified Drive folder.", "description": "All original images from the presentation are saved to the specified Drive folder.", "steps": [{"name": null, "description": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "text": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "At each original image location, the image has been removed (10pt).", "text": "At each original image location, the image has been removed (10pt).", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "text": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "text": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "text": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "text": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Each replacement image has a URL credit beneath it (10pt).", "text": "Each replacement image has a URL credit beneath it (10pt).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "text": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "text": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "text": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "description": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "steps": [{"name": null, "description": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "text": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "text": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 12, "total_points": 120, "requires_browsing_history": false, "source_assets": ["https://docs.google.com/presentation/d/1NAAZwHMwKjXt3A1BfZ4OJiUGWKaiv7Vx0c21sa1SW1E/edit"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_3/checkpoints.md"}
{"task_id": "knows.slides_17_remove_images_add_placeholders.4", "short_id": "17d", "family": "slides_17_removeimagesaddplaceholders", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Cross", "domain": "Slide Editing", "prompt": "Take the presentation at https://docs.google.com/presentation/d/1x_Ls6u6AFb7NkOVMkSDV5sPrAOCsb_DGOaz1nR85RmU/edit and help me replace all the images. First, make a copy of the original presentation and put in into {{OUTPUT_FOLDER_URL}}. Then identify all of the images in the document and save them all as image files and place them into {{IMAGES_FOLDER_URL}} for possible later use. Delete all the images, replacing each one with a description of the image in a text box at the same spot in big red text. Finally, for each of the images, find a similar image online and place it in the right spot. For each of these, make sure to put the URL for the image address below each new image to credit it.", "checkpoints": [{"n": 1, "title": "All original images from the presentation are saved to the specified Drive folder.", "description": "All original images from the presentation are saved to the specified Drive folder.", "steps": [{"name": null, "description": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "text": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "At each original image location, the image has been removed (10pt).", "text": "At each original image location, the image has been removed (10pt).", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "text": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "text": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "text": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "text": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Each replacement image has a URL credit beneath it (10pt).", "text": "Each replacement image has a URL credit beneath it (10pt).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "text": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "text": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "text": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "description": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "steps": [{"name": null, "description": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "text": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "text": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 12, "total_points": 120, "requires_browsing_history": false, "source_assets": ["https://docs.google.com/presentation/d/1x_Ls6u6AFb7NkOVMkSDV5sPrAOCsb_DGOaz1nR85RmU/edit"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_4/checkpoints.md"}
{"task_id": "knows.slides_17_remove_images_add_placeholders.5", "short_id": "17e", "family": "slides_17_removeimagesaddplaceholders", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Cross", "domain": "Slide Editing", "prompt": "Take the presentation at https://docs.google.com/presentation/d/1Gue5rJCgA0JMU1FRBo0TX8SdWl6ShZGisNm3o-y24TM/edit and help me replace all the images. First, make a copy of the original presentation and put in into {{OUTPUT_FOLDER_URL}}. Then identify all of the images in the document and save them all as image files and place them into {{IMAGES_FOLDER_URL}} for possible later use. Delete all the images, replacing each one with a description of the image in a text box at the same spot in big red text. Finally, for each of the images, find a similar image online and place it in the right spot. For each of these, make sure to put the URL for the image address below each new image to credit it.", "checkpoints": [{"n": 1, "title": "All original images from the presentation are saved to the specified Drive folder.", "description": "All original images from the presentation are saved to the specified Drive folder.", "steps": [{"name": null, "description": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "text": "All ground truth images from the original presentation are present in the Drive folder (verified via tiered image matching: exact -> perceptual hash -> VLM).", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "All images in the presentation are deleted and replaced with red text description placeholders. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "At each original image location, the image has been removed (10pt).", "text": "At each original image location, the image has been removed (10pt).", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "text": "A text box exists at the same location where each original image was - 80% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "text": "The text box contains a description of the original image - LLM similarity check against ground truth descriptions (10pt).", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "text": "The text in the placeholder is styled in big red text (>= 18pt) (10pt).", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "description": "Original images are replaced with similar images found online, with proper URL attribution. Points assigned based on percentage of images correctly processed, rounded to nearest 10%.", "steps": [{"name": null, "description": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "text": "Each image placeholder location now contains a replacement image from the web - 60% overlap required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Each replacement image has a URL credit beneath it (10pt).", "text": "Each replacement image has a URL credit beneath it (10pt).", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "text": "The URL below the image is a valid link to the image address online. The link must be below the image and in a valid location.(10pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "text": "VLM check confirms the replacement image is a reasonable substitute for the original image by comparing both images directly, with description-based fallback if image comparison fails. The image must be in the original location otherwise this test will fail as well. (10pt).", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "text": "The replacement image fully covers the text placeholder so no red text is visible - 90% coverage required (10pt).", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "description": "No extra images were added to the presentation beyond the replacements. No extra textboxes were added to the new presentation besides the the ones that go under the image and the textboxes that hold the links.", "steps": [{"name": null, "description": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "text": "The total number of images in the presentation matches the original count - no extra images were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct image count.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "text": "The total number of textboxes matches the orginal presentations count plus two times the original image count. No extra or less textboxes were added (10pt). Points are assigned proportionally based on the percentage of slides that have the correct textbox count.", "category": "structural_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 12, "total_points": 120, "requires_browsing_history": false, "source_assets": ["https://docs.google.com/presentation/d/1Gue5rJCgA0JMU1FRBo0TX8SdWl6ShZGisNm3o-y24TM/edit"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_17_removeimagesaddplaceholders/instance_5/checkpoints.md"}
{"task_id": "knows.slides_20_illustrated_book_report.1", "short_id": "20a", "family": "slides_20_Illustrated_Book_Report", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Education", "domain": "Book Report", "prompt": "I need to make a book report slideshow about the Red Rising Series book 1, can you help me get started? First, on the title slide put my name Stacey Johnson on the bottom, the title of the book above my name, and a photo of the cover of the book above that. Next, make a slide for at least 5 of the major characters with their name as the title and bulletpoints describing their characteristics below. At the bottom of each slide, include a section where you link to the source that you got the information for that character from. Make sure each character slide has a different background color. Finally, make a slide for the author of the book with their name and a photo of them. At the end, create a references slide and include links to at least 3 different sites that you used to create the presentations. Ensure that all information you enter in the slides is from these references.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "(1 pt) Exact match on name \"Stacey Johnson\"", "text": "(1 pt) Exact match on name \"Stacey Johnson\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Title of book present", "text": "(1 pt) Title of book present", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Photo of book cover present and valid", "text": "(1 pt) Photo of book cover present and valid", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "(1 pt) Structural Location match for name at bottom of slide, below title and photo.", "text": "(1 pt) Structural Location match for name at bottom of slide, below title and photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Structural Location match for title above name and below photo.", "text": "(1 pt) Structural Location match for title above name and below photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Photo is above both title and name.", "text": "(1 pt) Photo is above both title and name.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The character slides meet the requirements. Per character (5 characters, 24 pts each):", "description": "The character slides meet the requirements. Per character (5 characters, 24 pts each):", "steps": [{"name": null, "description": "(1 pt) The character chosen is among the top 20 characters from the book (based on a predefined list) and the character's name is the title.", "text": "(1 pt) The character chosen is among the top 20 characters from the book (based on a predefined list) and the character's name is the title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Each character slide has a different background color.", "text": "(1 pt) Each character slide has a different background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Each character slide has at least 3 bullet points with text.", "text": "(1 pt) Each character slide has at least 3 bullet points with text.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "(10 pt) Bullet points describe actual characteristics. Scored as floor(valid_count / total_count) * 10 where valid_count is the number of bullet points that describe characteristics.", "text": "(10 pt) Bullet points describe actual characteristics. Scored as floor(valid_count / total_count) * 10 where valid_count is the number of bullet points that describe characteristics.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "(1 pt) Each character slide has a source link at the bottom of the slide.", "text": "(1 pt) Each character slide has a source link at the bottom of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "(10 pt) The characteristics are supported by the source links on the slide. Scored as floor(validated_count / total_count) * 10 where validated_count is the number of characteristics corroborated by the cited source content.", "text": "(10 pt) The characteristics are supported by the source links on the slide. Scored as floor(validated_count / total_count) * 10 where validated_count is the number of characteristics corroborated by the cited source content.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The author slide is correct.", "description": "The author slide is correct.", "steps": [{"name": null, "description": "(1 pt) Exact match on author name.", "text": "(1 pt) Exact match on author name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) Photo of author present and valid. Em or VLM match with examples.", "text": "(1 pt) Photo of author present and valid. Em or VLM match with examples.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "References slide is correct.", "description": "References slide is correct.", "steps": [{"name": null, "description": "(1 pt) At least 3 different reference links present.", "text": "(1 pt) At least 3 different reference links present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "(1 pt) All reference links are valid URLs found in the agent trace.", "text": "(1 pt) All reference links are valid URLs found in the agent trace.", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 16, "total_points": 130, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_1/checkpoints.md"}
{"task_id": "knows.slides_20_illustrated_book_report.2", "short_id": "20b", "family": "slides_20_Illustrated_Book_Report", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Education", "domain": "Book Report", "prompt": "I need to make a book report slideshow about A Cat At the End of the World, can you help me get started? First, on the title slide put my name Anisija Mihaljević on the bottom, the title of the book above my name, and a photo of the cover of the book above that. Next, make a slide for at least 5 of the major characters with their name as the title and bulletpoints describing their characteristics below. At the bottom of each slide, include a section where you link to the source that you got the information for that character from. Make sure each character slide has a different background color. Finally, make a slide for the author of the book with their name and a photo of them. At the end, create a references slide and include links to at least 3 different sites that you used to create the presentations. Ensure that all information you enter in the slides is from these references.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on name \"Anisija Mihaljević\"", "text": "Exact match on name \"Anisija Mihaljević\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title of book present", "text": "Title of book present", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of book cover present and valid", "text": "Photo of book cover present and valid", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Structural Location match for name at bottom of slide, below title and photo.", "text": "Structural Location match for name at bottom of slide, below title and photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Structural Location match for title above name and below photo.", "text": "Structural Location match for title above name and below photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Photo is above both title and name.", "text": "Photo is above both title and name.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The character slides meet the requirements.", "description": "The character slides meet the requirements.", "steps": [{"name": null, "description": "The character chosen is among the a list of the top 20 characters from the book (based on a predefined list) and the character's name is the title.", "text": "The character chosen is among the a list of the top 20 characters from the book (based on a predefined list) and the character's name is the title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a different background color.", "text": "Each character slide has a different background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has at least 3 bullet points describing their characteristics.", "text": "Each character slide has at least 3 bullet points describing their characteristics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a source link at the bottom of the slide.", "text": "Each character slide has a source link at the bottom of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The characteristics are all direct quotes from the sources links in the slide.", "text": "The characteristics are all direct quotes from the sources links in the slide.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The author slide is correct.", "description": "The author slide is correct.", "steps": [{"name": null, "description": "Exact match on author name.", "text": "Exact match on author name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of author present and valid. Em or VLM match with examples.", "text": "Photo of author present and valid. Em or VLM match with examples.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "References slide is correct.", "description": "References slide is correct.", "steps": [{"name": null, "description": "At least 3 different reference links present.", "text": "At least 3 different reference links present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All reference links are valid URLs found in the agent trace.", "text": "All reference links are valid URLs found in the agent trace.", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 15, "total_points": 35, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_2/checkpoints.md"}
{"task_id": "knows.slides_20_illustrated_book_report.3", "short_id": "20c", "family": "slides_20_Illustrated_Book_Report", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Education", "domain": "Book Report", "prompt": "I need to make a book report slideshow about Jane Eyre, can you help me get started? First, on the title slide put my name Jane Austen on the bottom, the title of the book above my name, and a photo of the cover of the book above that. Next, make a slide for at least 5 of the major characters with their name as the title and bulletpoints describing their characteristics below. At the bottom of each slide, include a section where you link to the source that you got the information for that character from. Make sure each character slide has a different background color. Finally, make a slide for the author of the book with their name and a photo of them. At the end, create a references slide and include links to at least 3 different sites that you used to create the presentations. Ensure that all information you enter in the slides is from these references.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on name \"Jane Austen\"", "text": "Exact match on name \"Jane Austen\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title of book present", "text": "Title of book present", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of book cover present and valid", "text": "Photo of book cover present and valid", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Structural Location match for name at bottom of slide, below title and photo.", "text": "Structural Location match for name at bottom of slide, below title and photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Structural Location match for title above name and below photo.", "text": "Structural Location match for title above name and below photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Photo is above both title and name.", "text": "Photo is above both title and name.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The character slides meet the requirements.", "description": "The character slides meet the requirements.", "steps": [{"name": null, "description": "The character chosen is among the list of major characters from the book (based on a predefined list of 27 characters with alias groups) and the character's name is the title. Character-and-alias pairs (e.g., \"Jane Eyre\" and \"the governess\", \"Bertha Mason\" and \"the madwoman\") are treated as the same character and cannot both occupy character-slide slots.", "text": "The character chosen is among the list of major characters from the book (based on a predefined list of 27 characters with alias groups) and the character's name is the title. Character-and-alias pairs (e.g., \"Jane Eyre\" and \"the governess\", \"Bertha Mason\" and \"the madwoman\") are treated as the same character and cannot both occupy character-slide slots.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a different background color.", "text": "Each character slide has a different background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has at least 3 bullet points describing their characteristics.", "text": "Each character slide has at least 3 bullet points describing their characteristics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a source link at the bottom of the slide.", "text": "Each character slide has a source link at the bottom of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The characteristics are all direct quotes from the sources links in the slide.", "text": "The characteristics are all direct quotes from the sources links in the slide.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The author slide is correct.", "description": "The author slide is correct.", "steps": [{"name": null, "description": "Exact match on author name: either \"Charlotte Brontë\" (diacritics required) or the pseudonym \"Currer Bell\" under which Jane Eyre was first published.", "text": "Exact match on author name: either \"Charlotte Brontë\" (diacritics required) or the pseudonym \"Currer Bell\" under which Jane Eyre was first published.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of author present and valid. Em or VLM match with examples.", "text": "Photo of author present and valid. Em or VLM match with examples.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "References slide is correct.", "description": "References slide is correct.", "steps": [{"name": null, "description": "At least 3 different reference links present.", "text": "At least 3 different reference links present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All reference links are valid URLs found in the agent trace.", "text": "All reference links are valid URLs found in the agent trace.", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 15, "total_points": 35, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_3/checkpoints.md"}
{"task_id": "knows.slides_20_illustrated_book_report.4", "short_id": "20d", "family": "slides_20_Illustrated_Book_Report", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Education", "domain": "Book Report", "prompt": "I need to make a book report slideshow about The Eyre Affair by Jasper Fforde, can you help me get started? First, on the title slide put my name Charlotte Currer Bell on the bottom, the title of the book above my name, and a photo of the cover of the book above that. Next, make a slide for at least 5 of the major characters with their name as the title and bulletpoints describing their characteristics below. At the bottom of each slide, include a section where you link to the source that you got the information for that character from. Make sure each character slide has a different background color. Finally, make a slide for the author of the book with their name and a photo of them. At the end, create a references slide and include links to at least 3 different sites that you used to create the presentations. Ensure that all information you enter in the slides is from these references.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on name \"Charlotte Currer Bell\"", "text": "Exact match on name \"Charlotte Currer Bell\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title of book present", "text": "Title of book present", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of book cover present and valid", "text": "Photo of book cover present and valid", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Structural Location match for name at bottom of slide, below title and photo.", "text": "Structural Location match for name at bottom of slide, below title and photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Structural Location match for title above name and below photo.", "text": "Structural Location match for title above name and below photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Photo is above both title and name.", "text": "Photo is above both title and name.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The character slides meet the requirements.", "description": "The character slides meet the requirements.", "steps": [{"name": null, "description": "The character chosen is among the list of major characters from the book (based on a predefined list of characters with alias groups, derived from the Wikipedia article for The Eyre Affair) and the character's name is the title. Character-and-alias pairs (e.g., \"Thursday Next\" and \"Thursday\") are treated as the same character and cannot both occupy character-slide slots.", "text": "The character chosen is among the list of major characters from the book (based on a predefined list of characters with alias groups, derived from the Wikipedia article for The Eyre Affair) and the character's name is the title. Character-and-alias pairs (e.g., \"Thursday Next\" and \"Thursday\") are treated as the same character and cannot both occupy character-slide slots.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a different background color.", "text": "Each character slide has a different background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has at least 3 bullet points describing their characteristics.", "text": "Each character slide has at least 3 bullet points describing their characteristics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a source link at the bottom of the slide.", "text": "Each character slide has a source link at the bottom of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The characteristics are all direct quotes from the sources links in the slide.", "text": "The characteristics are all direct quotes from the sources links in the slide.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The author slide is correct.", "description": "The author slide is correct.", "steps": [{"name": null, "description": "Exact match on author name \"Jasper Fforde\".", "text": "Exact match on author name \"Jasper Fforde\".", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of author present and valid. Em or VLM match with examples.", "text": "Photo of author present and valid. Em or VLM match with examples.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "References slide is correct.", "description": "References slide is correct.", "steps": [{"name": null, "description": "At least 3 different reference links present.", "text": "At least 3 different reference links present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All reference links are valid URLs found in the agent trace.", "text": "All reference links are valid URLs found in the agent trace.", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 15, "total_points": 35, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_4/checkpoints.md"}
{"task_id": "knows.slides_20_illustrated_book_report.5", "short_id": "20e", "family": "slides_20_Illustrated_Book_Report", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Education", "domain": "Book Report", "prompt": "I need to make a book report slideshow about Slade House by David Mitchell, can you help me get started? First, on the title slide put my name John Green on the bottom, the title of the book above my name, and a photo of the cover of the book above that. Next, make a slide for at least 5 of the major characters with their name as the title and bulletpoints describing their characteristics below. At the bottom of each slide, include a section where you link to the source that you got the information for that character from. Make sure each character slide has a different background color. Finally, make a slide for the author of the book with their name and a photo of them. At the end, create a references slide and include links to at least 3 different sites that you used to create the presentations. Ensure that all information you enter in the slides is from these references.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on name \"John Green\"", "text": "Exact match on name \"John Green\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title of book present", "text": "Title of book present", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of book cover present and valid", "text": "Photo of book cover present and valid", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Structural Location match for name at bottom of slide, below title and photo.", "text": "Structural Location match for name at bottom of slide, below title and photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Structural Location match for title above name and below photo.", "text": "Structural Location match for title above name and below photo.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Photo is above both title and name.", "text": "Photo is above both title and name.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The character slides meet the requirements.", "description": "The character slides meet the requirements.", "steps": [{"name": null, "description": "The character chosen is among the list of major characters from the book (based on a predefined list of characters with alias groups, derived from the Wikipedia article for Slade House) and the character's name is the title. Character-and-alias pairs (e.g., \"Mrs Bishop\" and \"Rita Bishop\", \"Iris Marinus-Fenby\" and \"Marinus\") are treated as the same character and cannot both occupy character-slide slots.", "text": "The character chosen is among the list of major characters from the book (based on a predefined list of characters with alias groups, derived from the Wikipedia article for Slade House) and the character's name is the title. Character-and-alias pairs (e.g., \"Mrs Bishop\" and \"Rita Bishop\", \"Iris Marinus-Fenby\" and \"Marinus\") are treated as the same character and cannot both occupy character-slide slots.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a different background color.", "text": "Each character slide has a different background color.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has at least 3 bullet points describing their characteristics.", "text": "Each character slide has at least 3 bullet points describing their characteristics.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Each character slide has a source link at the bottom of the slide.", "text": "Each character slide has a source link at the bottom of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The characteristics are all direct quotes from the sources links in the slide.", "text": "The characteristics are all direct quotes from the sources links in the slide.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "The author slide is correct.", "description": "The author slide is correct.", "steps": [{"name": null, "description": "Exact match on author name \"David Mitchell\".", "text": "Exact match on author name \"David Mitchell\".", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Photo of author present and valid. Em or VLM match with examples.", "text": "Photo of author present and valid. Em or VLM match with examples.", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "References slide is correct.", "description": "References slide is correct.", "steps": [{"name": null, "description": "At least 3 different reference links present.", "text": "At least 3 different reference links present.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All reference links are valid URLs found in the agent trace.", "text": "All reference links are valid URLs found in the agent trace.", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 15, "total_points": 35, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_20_Illustrated_Book_Report/instance_5/checkpoints.md"}
{"task_id": "knows.slides_26_basic_educational_slide_deck.1", "short_id": "26a", "family": "slides_26_basic_educational_slide_deck", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Education", "domain": "Lesson Plan", "prompt": "I'm preparing an introductory teaching presentation on 'the History of Video Games' for a 7th-grade audience, and I'd like help designing the slide content from scratch.\n\nStart with a title slide that highlights the History of Video Games in large, bold, dark-orange lettering, includes my name (Tina Smith), and uses a thematically relevant background image. Add a small-font source credit beneath the image.\n\nNext, break the History of Video Games into five student-friendly sections. For each section, locate one reliable online reference through web search, and create a slide that includes:\n- A bold-italic section heading\n- A short list of bullet-point explanations based on the source you found\n- A clear citation for the source in small type in the lower-left corner\n- An illustrative online image, placed on the right side, with its source credit beneath it in small text.\n\nFinish with a slide that summarizes the key ideas and encourages the student to ask questions.", "checkpoints": [{"n": 1, "title": "Title Slide", "description": "Verify that the title slide contains the presentation topic in bold dark orange font, the presenter’s name, a relevant background image, and proper image attribution.", "steps": [{"name": null, "description": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "text": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Topic is in dark orange (2 pt)", "text": "Topic is in dark orange (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Includes the presenter’s name (3 pt)", "text": "Includes the presenter’s name (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title slide has a background image and it is relevant to the topic (5 pt)", "text": "Title slide has a background image and it is relevant to the topic (5 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image credit is clearly shown in small font (2 pt)", "text": "Image credit is clearly shown in small font (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Content Slides", "description": "Verify that each content slide covers a distinct subsection of the topic with a properly formatted heading, sourced bullet-point content, and correct layout. Steps 2-10 award partial credit proportional to the number of passing slides out of the actual content slide count.", "steps": [{"name": null, "description": "Exactly N content slides (parsed from task.md) (5 pt)", "text": "Exactly N content slides (parsed from task.md) (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All content slides are related to the topic (10 pt)", "text": "All content slides are related to the topic (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "text": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The information in the slide is actually coming from the cited source (10 pt)", "text": "The information in the slide is actually coming from the cited source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each slide has a unique, clear heading related to its subsection (10 pt)", "text": "Each slide has a unique, clear heading related to its subsection (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Heading is bold and italic (10 pt)", "text": "Heading is bold and italic (10 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Information is presented in bullet points (5 pt)", "text": "Information is presented in bullet points (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No text box on the slide overflows the slide bounds (5 pt)", "text": "No text box on the slide overflows the slide bounds (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "text": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The content is placed on the left side of the slide (10 pt)", "text": "The content is placed on the left side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Summary Slide", "description": "Verify that the final slide summarizes all covered content in original language and includes a prompt for student engagement.", "steps": [{"name": null, "description": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "text": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary text is original, not verbatim from content slides or sources (3 pt)", "text": "Summary text is original, not verbatim from content slides or sources (3 pt)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "text": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Visual Elements", "description": "Verify that each content slide has a relevant image placed on the right side with a valid, attributed source link. Each step awards partial credit proportional to the number of passing slides.", "steps": [{"name": null, "description": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "text": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is placed on the right side of the slide (10 pt)", "text": "Image is placed on the right side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "text": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Images are unique (10 pt, proportional)", "text": "The Images are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 24, "total_points": 165, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_1/checkpoints.md"}
{"task_id": "knows.slides_26_basic_educational_slide_deck.2", "short_id": "26b", "family": "slides_26_basic_educational_slide_deck", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Education", "domain": "Lesson Plan", "prompt": "I'm preparing an introductory teaching presentation on 'CRISPR-Cas9 Gene Editing' for a 10th-grade audience, and I'd like help designing the slide content from scratch.\n\nStart with a title slide that highlights the CRISPR-Cas9 Gene Editing in large, bold, dark-orange lettering, includes my name (Mr. Daniel Lee), and uses a thematically relevant background image. Add a small-font source credit beneath the image.\n\nNext, break the CRISPR-Cas9 Gene Editing into five student-friendly sections. For each section, locate one reliable online reference through web search, and create a slide that includes:\n- A bold-italic section heading\n- A short list of bullet-point explanations based on the source you found\n- A clear citation for the source in small type in the lower-left corner\n- An illustrative online image, placed on the right side, with its source credit beneath it in small text.\n\nFinish with a slide that summarizes the key ideas and encourages the student to ask questions.", "checkpoints": [{"n": 1, "title": "Title Slide", "description": "Verify that the title slide contains the presentation topic in bold dark orange font, the presenter’s name, a relevant background image, and proper image attribution.", "steps": [{"name": null, "description": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "text": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Topic is in dark orange (2 pt)", "text": "Topic is in dark orange (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Includes the presenter’s name (3 pt)", "text": "Includes the presenter’s name (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title slide has a background image and it is relevant to the topic (5 pt)", "text": "Title slide has a background image and it is relevant to the topic (5 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image credit is clearly shown in small font (2 pt)", "text": "Image credit is clearly shown in small font (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Content Slides", "description": "Verify that each content slide covers a distinct subsection of the topic with a properly formatted heading, sourced bullet-point content, and correct layout. Steps 2-10 award partial credit proportional to the number of passing slides out of the actual content slide count.", "steps": [{"name": null, "description": "Exactly N content slides (parsed from task.md) (5 pt)", "text": "Exactly N content slides (parsed from task.md) (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All content slides are related to the topic (10 pt)", "text": "All content slides are related to the topic (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "text": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The information in the slide is actually coming from the cited source (10 pt)", "text": "The information in the slide is actually coming from the cited source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each slide has a unique, clear heading related to its subsection (10 pt)", "text": "Each slide has a unique, clear heading related to its subsection (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Heading is bold and italic (10 pt)", "text": "Heading is bold and italic (10 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Information is presented in bullet points (5 pt)", "text": "Information is presented in bullet points (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No text box on the slide overflows the slide bounds (5 pt)", "text": "No text box on the slide overflows the slide bounds (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "text": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The content is placed on the left side of the slide (10 pt)", "text": "The content is placed on the left side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Summary Slide", "description": "Verify that the final slide summarizes all covered content in original language and includes a prompt for student engagement.", "steps": [{"name": null, "description": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "text": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary text is original, not verbatim from content slides or sources (3 pt)", "text": "Summary text is original, not verbatim from content slides or sources (3 pt)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "text": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Visual Elements", "description": "Verify that each content slide has a relevant image placed on the right side with a valid, attributed source link. Each step awards partial credit proportional to the number of passing slides.", "steps": [{"name": null, "description": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "text": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is placed on the right side of the slide (10 pt)", "text": "Image is placed on the right side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "text": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Images are unique (10 pt, proportional)", "text": "The Images are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 24, "total_points": 165, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_2/checkpoints.md"}
{"task_id": "knows.slides_26_basic_educational_slide_deck.3", "short_id": "26c", "family": "slides_26_basic_educational_slide_deck", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Education", "domain": "Lesson Plan", "prompt": "I'm preparing an introductory teaching presentation on 'the Seven Wonders of the Ancient World' for a 6th-grade audience, and I'd like help designing the slide content from scratch.\n\nStart with a title slide that highlights the Seven Wonders of the Ancient World in large, bold, dark-orange lettering, includes my name (Sarah Johnson), and uses a thematically relevant background image. Add a small-font source credit beneath the image.\n\nNext, break the Seven Wonders of the Ancient World into seven student-friendly sections. For each section, locate one reliable online reference through web search, and create a slide that includes:\n- A bold-italic section heading\n- A short list of bullet-point explanations based on the source you found\n- A clear citation for the source in small type in the lower-left corner\n- An illustrative online image, placed on the right side, with its source credit beneath it in small text.\n\nFinish with a slide that summarizes the key ideas and encourages the student to ask questions.", "checkpoints": [{"n": 1, "title": "Title Slide", "description": "Verify that the title slide contains the presentation topic in bold dark orange font, the presenter’s name, a relevant background image, and proper image attribution.", "steps": [{"name": null, "description": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "text": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Topic is in dark orange (2 pt)", "text": "Topic is in dark orange (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Includes the presenter’s name (3 pt)", "text": "Includes the presenter’s name (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title slide has a background image and it is relevant to the topic (5 pt)", "text": "Title slide has a background image and it is relevant to the topic (5 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image credit is clearly shown in small font (2 pt)", "text": "Image credit is clearly shown in small font (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Content Slides", "description": "Verify that each content slide covers a distinct subsection of the topic with a properly formatted heading, sourced bullet-point content, and correct layout. Steps 2-10 award partial credit proportional to the number of passing slides out of the actual content slide count.", "steps": [{"name": null, "description": "Exactly N content slides (parsed from task.md) (5 pt)", "text": "Exactly N content slides (parsed from task.md) (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All content slides are related to the topic (10 pt)", "text": "All content slides are related to the topic (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "text": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The information in the slide is actually coming from the cited source (10 pt)", "text": "The information in the slide is actually coming from the cited source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each slide has a unique, clear heading related to its subsection (10 pt)", "text": "Each slide has a unique, clear heading related to its subsection (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Heading is bold and italic (10 pt)", "text": "Heading is bold and italic (10 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Information is presented in bullet points (5 pt)", "text": "Information is presented in bullet points (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No text box on the slide overflows the slide bounds (5 pt)", "text": "No text box on the slide overflows the slide bounds (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "text": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The content is placed on the left side of the slide (10 pt)", "text": "The content is placed on the left side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Summary Slide", "description": "Verify that the final slide summarizes all covered content in original language and includes a prompt for student engagement.", "steps": [{"name": null, "description": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "text": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary text is original, not verbatim from content slides or sources (3 pt)", "text": "Summary text is original, not verbatim from content slides or sources (3 pt)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "text": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Visual Elements", "description": "Verify that each content slide has a relevant image placed on the right side with a valid, attributed source link. Each step awards partial credit proportional to the number of passing slides.", "steps": [{"name": null, "description": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "text": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is placed on the right side of the slide (10 pt)", "text": "Image is placed on the right side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "text": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Images are unique (10 pt, proportional)", "text": "The Images are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 24, "total_points": 165, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_3/checkpoints.md"}
{"task_id": "knows.slides_26_basic_educational_slide_deck.4", "short_id": "26d", "family": "slides_26_basic_educational_slide_deck", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Education", "domain": "Lesson Plan", "prompt": "I'm preparing an introductory teaching presentation on 'the History of K-Pop' for a 9th-grade audience, and I'd like help designing the slide content from scratch.\n\nStart with a title slide that highlights the History of K-Pop in large, bold, dark-orange lettering, includes my name (David Martinez), and uses a thematically relevant background image. Add a small-font source credit beneath the image.\n\nNext, break the History of K-Pop into six student-friendly sections. For each section, locate one reliable online reference through web search, and create a slide that includes:\n- A bold-italic section heading\n- A short list of bullet-point explanations based on the source you found\n- A clear citation for the source in small type in the lower-left corner\n- An illustrative online image, placed on the right side, with its source credit beneath it in small text.\n\nFinish with a slide that summarizes the key ideas and encourages the student to ask questions.", "checkpoints": [{"n": 1, "title": "Title Slide", "description": "Verify that the title slide contains the presentation topic in bold dark orange font, the presenter’s name, a relevant background image, and proper image attribution.", "steps": [{"name": null, "description": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "text": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Topic is in dark orange (2 pt)", "text": "Topic is in dark orange (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Includes the presenter’s name (3 pt)", "text": "Includes the presenter’s name (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title slide has a background image and it is relevant to the topic (5 pt)", "text": "Title slide has a background image and it is relevant to the topic (5 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image credit is clearly shown in small font (2 pt)", "text": "Image credit is clearly shown in small font (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Content Slides", "description": "Verify that each content slide covers a distinct subsection of the topic with a properly formatted heading, sourced bullet-point content, and correct layout. Steps 2-10 award partial credit proportional to the number of passing slides out of the actual content slide count.", "steps": [{"name": null, "description": "Exactly N content slides (parsed from task.md) (5 pt)", "text": "Exactly N content slides (parsed from task.md) (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All content slides are related to the topic (10 pt)", "text": "All content slides are related to the topic (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "text": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The information in the slide is actually coming from the cited source (10 pt)", "text": "The information in the slide is actually coming from the cited source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each slide has a unique, clear heading related to its subsection (10 pt)", "text": "Each slide has a unique, clear heading related to its subsection (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Heading is bold and italic (10 pt)", "text": "Heading is bold and italic (10 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Information is presented in bullet points (5 pt)", "text": "Information is presented in bullet points (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No text box on the slide overflows the slide bounds (5 pt)", "text": "No text box on the slide overflows the slide bounds (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "text": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The content is placed on the left side of the slide (10 pt)", "text": "The content is placed on the left side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Summary Slide", "description": "Verify that the final slide summarizes all covered content in original language and includes a prompt for student engagement.", "steps": [{"name": null, "description": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "text": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary text is original, not verbatim from content slides or sources (3 pt)", "text": "Summary text is original, not verbatim from content slides or sources (3 pt)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "text": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Visual Elements", "description": "Verify that each content slide has a relevant image placed on the right side with a valid, attributed source link. Each step awards partial credit proportional to the number of passing slides.", "steps": [{"name": null, "description": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "text": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is placed on the right side of the slide (10 pt)", "text": "Image is placed on the right side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "text": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Images are unique (10 pt, proportional)", "text": "The Images are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 24, "total_points": 165, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_4/checkpoints.md"}
{"task_id": "knows.slides_26_basic_educational_slide_deck.5", "short_id": "26e", "family": "slides_26_basic_educational_slide_deck", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Education", "domain": "Lesson Plan", "prompt": "I'm preparing an introductory teaching presentation on 'Mercury' for a 7th-grade audience, and I'd like help designing the slide content from scratch.\n\nStart with a title slide that highlights the Mercury in large, bold, dark-orange lettering, includes my name (Mrs. Emily Carter), and uses a thematically relevant background image. Add a small-font source credit beneath the image.\n\nNext, break the Mercury into four student-friendly sections. For each section, locate one reliable online reference through web search, and create a slide that includes:\n- A bold-italic section heading\n- A short list of bullet-point explanations based on the source you found\n- A clear citation for the source in small type in the lower-left corner\n- An illustrative online image, placed on the right side, with its source credit beneath it in small text.\n\nFinish with a slide that summarizes the key ideas and encourages the student to ask questions.", "checkpoints": [{"n": 1, "title": "Title Slide", "description": "Verify that the title slide contains the presentation topic in bold dark orange font, the presenter’s name, a relevant background image, and proper image attribution.", "steps": [{"name": null, "description": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "text": "Contains the topic in big, bold font (font size ≥ 40pt) (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Topic is in dark orange (2 pt)", "text": "Topic is in dark orange (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Includes the presenter’s name (3 pt)", "text": "Includes the presenter’s name (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title slide has a background image and it is relevant to the topic (5 pt)", "text": "Title slide has a background image and it is relevant to the topic (5 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image credit is clearly shown in small font (2 pt)", "text": "Image credit is clearly shown in small font (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (3 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Content Slides", "description": "Verify that each content slide covers a distinct subsection of the topic with a properly formatted heading, sourced bullet-point content, and correct layout. Steps 2-10 award partial credit proportional to the number of passing slides out of the actual content slide count.", "steps": [{"name": null, "description": "Exactly N content slides (parsed from task.md) (5 pt)", "text": "Exactly N content slides (parsed from task.md) (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All content slides are related to the topic (10 pt)", "text": "All content slides are related to the topic (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "text": "Each subsection has a valid source of information in small font and the source is placed in the lower-left corner (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The information in the slide is actually coming from the cited source (10 pt)", "text": "The information in the slide is actually coming from the cited source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Each slide has a unique, clear heading related to its subsection (10 pt)", "text": "Each slide has a unique, clear heading related to its subsection (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Heading is bold and italic (10 pt)", "text": "Heading is bold and italic (10 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Information is presented in bullet points (5 pt)", "text": "Information is presented in bullet points (5 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "No text box on the slide overflows the slide bounds (5 pt)", "text": "No text box on the slide overflows the slide bounds (5 pt)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "text": "Content is paraphrased/summarized rather than verbatim copied from the source (10 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The content is placed on the left side of the slide (10 pt)", "text": "The content is placed on the left side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Summary Slide", "description": "Verify that the final slide summarizes all covered content in original language and includes a prompt for student engagement.", "steps": [{"name": null, "description": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "text": "Summary slide briefly mentions all of the concepts covered in the previous slides (4 pt)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary text is original, not verbatim from content slides or sources (3 pt)", "text": "Summary text is original, not verbatim from content slides or sources (3 pt)", "category": "content_check", "llm_vlm_primary": false}, {"name": null, "description": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "text": "Slide includes a statement or prompt encouraging students to ask questions or engage (3 pt)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Visual Elements", "description": "Verify that each content slide has a relevant image placed on the right side with a valid, attributed source link. Each step awards partial credit proportional to the number of passing slides.", "steps": [{"name": null, "description": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "text": "Each content slide includes one image that is directly relevant to the subsection, excluding the title slide and summary slide (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is placed on the right side of the slide (10 pt)", "text": "Image is placed on the right side of the slide (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "text": "Every image has a source credit in small font beneath it on the lower-right side (10 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "text": "Image source URL is accessible and points to the same image or a page containing the image (10 pt)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Images are unique (10 pt, proportional)", "text": "The Images are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 24, "total_points": 165, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_26_basic_educational_slide_deck/instance_5/checkpoints.md"}
{"task_id": "knows.slides_29_buy_car_pres.1", "short_id": "29a", "family": "slides_29_buy_car_pres", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Personal", "domain": "Purchase Decision", "prompt": "Create a slides presentation comparing different cars that can be categorized as a minivan. First, find an online article talking about at least 5 different 2023 models. Make a title slide called \"Comparing Different Cool Cars to Buy\" with font size atleast 30pt. Then for each of at least 5 cars of that category, make a slide for each. Start with the make and model as the slide title. Then, for each car, go to Kelly Blue Book and find the sticker price, fuel efficiency and horsepower. Then, go to Kelly Blue Book for user reviews and get the average rating for each. Include the link to the review website in the slide. For each car, add a picture of the car that takes up at least 50% of the slide. The picture must be from the article or Kelly Blue Book. Finally, on the last slide, list which car had the best stats in each category (sticker price, fuel efficiency, horsepower, user rating).", "checkpoints": [{"n": 1, "title": "Title slide contains the correct text.", "description": "Title slide contains the correct text.", "steps": [{"name": null, "description": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "text": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matched text has a font size of at least 30pt.", "text": "The matched text has a font size of at least 30pt.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Car content slides exists and have correct cars.", "description": "Car content slides exists and have correct cars.", "steps": [{"name": null, "description": "The browsing history contains a visit to the article talking about 2023 best minivans.", "text": "The browsing history contains a visit to the article talking about 2023 best minivans.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "text": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "category": "structural_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Car slides meet the requirements.", "description": "Car slides meet the requirements.", "steps": [{"name": null, "description": "The browsing history contains a visit to the corresponding KBB vehicle page.", "text": "The browsing history contains a visit to the corresponding KBB vehicle page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The make and model of the car are listed in the slide title.", "text": "The make and model of the car are listed in the slide title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A picture of the correct model is found.", "text": "A picture of the correct model is found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The car picture takes up at least 50% of the slide.", "text": "The car picture takes up at least 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "text": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Sticker price matches the price listed on Kelly Blue Book.", "text": "Sticker price matches the price listed on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "text": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Horsepower stat matches the listed value on Kelly Blue Book.", "text": "Horsepower stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a user review platform is provided.", "text": "A URL to a user review platform is provided.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the user review platform URL.", "text": "The browsing history contains a visit to the user review platform URL.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "User average rating matches the listed value on the user review platform.", "text": "User average rating matches the listed value on the user review platform.", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 4, "title": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "description": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "steps": [{"name": null, "description": "The slide title denotes that the slide contains the best car stats.", "text": "The slide title denotes that the slide contains the best car stats.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The correct lowest-price car is listed.", "text": "The correct lowest-price car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest MPG car is listed.", "text": "The correct highest MPG car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest horsepower car is listed.", "text": "The correct highest horsepower car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The most highly-rated car is listed.", "text": "The most highly-rated car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_1/checkpoints.md"}
{"task_id": "knows.slides_29_buy_car_pres.2", "short_id": "29b", "family": "slides_29_buy_car_pres", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Personal", "domain": "Purchase Decision", "prompt": "Create a slides presentation comparing different cars that can be categorized as a diesel sedan. First, find an online article talking about at least 5 different 2015 models. Make a title slide called \"Comparing Different Cool Cars to Buy\" with font size atleast 30pt. Then for each of at least 5 cars of that category, make a slide for each. Start with the make and model as the slide title. Then, for each car, go to Kelly Blue Book and find the sticker price, fuel efficiency and horsepower. Then, go to Kelly Blue Book for user reviews and get the average rating for each. Include the link to the review website in the slide. For each car, add a picture of the car that takes up at least 50% of the slide. The picture must be from the article or Kelly Blue Book. Finally, on the last slide, list which car had the best stats in each category (sticker price, fuel efficiency, horsepower, user rating).", "checkpoints": [{"n": 1, "title": "Title slide contains the correct text.", "description": "Title slide contains the correct text.", "steps": [{"name": null, "description": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "text": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matched text has a font size of at least 30pt.", "text": "The matched text has a font size of at least 30pt.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Car content slides exists and have correct cars.", "description": "Car content slides exists and have correct cars.", "steps": [{"name": null, "description": "The browsing history contains a visit to the article talking about 2015 best diesel sedans.", "text": "The browsing history contains a visit to the article talking about 2015 best diesel sedans.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "text": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "category": "structural_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Car slides meet the requirements.", "description": "Car slides meet the requirements.", "steps": [{"name": null, "description": "The browsing history contains a visit to the corresponding KBB vehicle page.", "text": "The browsing history contains a visit to the corresponding KBB vehicle page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The make and model of the car are listed in the slide title.", "text": "The make and model of the car are listed in the slide title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A picture of the correct model is found.", "text": "A picture of the correct model is found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The car picture takes up at least 50% of the slide.", "text": "The car picture takes up at least 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "text": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Sticker price matches the price listed on Kelly Blue Book.", "text": "Sticker price matches the price listed on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "text": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Horsepower stat matches the listed value on Kelly Blue Book.", "text": "Horsepower stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a user review platform is provided.", "text": "A URL to a user review platform is provided.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the user review platform URL.", "text": "The browsing history contains a visit to the user review platform URL.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "User average rating matches the listed value on the user review platform.", "text": "User average rating matches the listed value on the user review platform.", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 4, "title": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "description": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "steps": [{"name": null, "description": "The slide title denotes that the slide contains the best car stats.", "text": "The slide title denotes that the slide contains the best car stats.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The correct lowest-price car is listed.", "text": "The correct lowest-price car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest MPG car is listed.", "text": "The correct highest MPG car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest horsepower car is listed.", "text": "The correct highest horsepower car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The most highly-rated car is listed.", "text": "The most highly-rated car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_2/checkpoints.md"}
{"task_id": "knows.slides_29_buy_car_pres.3", "short_id": "29c", "family": "slides_29_buy_car_pres", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Personal", "domain": "Purchase Decision", "prompt": "Create a slides presentation comparing different cars that can be categorized as an electric pickup truck. First, find an online article talking about at least 5 different 2026 models. Make a title slide called \"Comparing Different Cool Cars to Buy\" with font size atleast 30pt. Then for each of at least 5 cars of that category, make a slide for each. Start with the make and model as the slide title. Then, for each car, go to Kelly Blue Book and find the sticker price, fuel efficiency and horsepower. Then, go to Kelly Blue Book for user reviews and get the average rating for each. Include the link to the review website in the slide. For each car, add a picture of the car that takes up at least 50% of the slide. The picture must be from the article or Kelly Blue Book. Finally, on the last slide, list which car had the best stats in each category (sticker price, fuel efficiency, horsepower, user rating).", "checkpoints": [{"n": 1, "title": "Title slide contains the correct text.", "description": "Title slide contains the correct text.", "steps": [{"name": null, "description": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "text": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matched text has a font size of at least 30pt.", "text": "The matched text has a font size of at least 30pt.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Car content slides exists and have correct cars.", "description": "Car content slides exists and have correct cars.", "steps": [{"name": null, "description": "The browsing history contains a visit to the article talking about 2026 best electric pickup trucks.", "text": "The browsing history contains a visit to the article talking about 2026 best electric pickup trucks.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "text": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "category": "structural_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Car slides meet the requirements.", "description": "Car slides meet the requirements.", "steps": [{"name": null, "description": "The browsing history contains a visit to the corresponding KBB vehicle page.", "text": "The browsing history contains a visit to the corresponding KBB vehicle page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The make and model of the car are listed in the slide title.", "text": "The make and model of the car are listed in the slide title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A picture of the correct model is found.", "text": "A picture of the correct model is found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The car picture takes up at least 50% of the slide.", "text": "The car picture takes up at least 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "text": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Sticker price matches the price listed on Kelly Blue Book.", "text": "Sticker price matches the price listed on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "text": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Horsepower stat matches the listed value on Kelly Blue Book.", "text": "Horsepower stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a user review platform is provided.", "text": "A URL to a user review platform is provided.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the user review platform URL.", "text": "The browsing history contains a visit to the user review platform URL.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "User average rating matches the listed value on the user review platform.", "text": "User average rating matches the listed value on the user review platform.", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 4, "title": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "description": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "steps": [{"name": null, "description": "The slide title denotes that the slide contains the best car stats.", "text": "The slide title denotes that the slide contains the best car stats.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The correct lowest-price car is listed.", "text": "The correct lowest-price car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest MPG car is listed.", "text": "The correct highest MPG car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest horsepower car is listed.", "text": "The correct highest horsepower car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The most highly-rated car is listed.", "text": "The most highly-rated car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_3/checkpoints.md"}
{"task_id": "knows.slides_29_buy_car_pres.4", "short_id": "29d", "family": "slides_29_buy_car_pres", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Personal", "domain": "Purchase Decision", "prompt": "Create a slides presentation comparing different cars that can be categorized as a coupe. First, find an online article talking about at least 5 different 2005 models. Make a title slide called \"Comparing Different Cool Cars to Buy\" with font size atleast 30pt. Then for each of at least 5 cars of that category, make a slide for each. Start with the make and model as the slide title. Then, for each car, go to Kelly Blue Book and find the sticker price, fuel efficiency and horsepower. Then, go to Kelly Blue Book for user reviews and get the average rating for each. Include the link to the review website in the slide. For each car, add a picture of the car that takes up at least 50% of the slide. The picture must be from the article or Kelly Blue Book. Finally, on the last slide, list which car had the best stats in each category (sticker price, fuel efficiency, horsepower, user rating).", "checkpoints": [{"n": 1, "title": "Title slide contains the correct text.", "description": "Title slide contains the correct text.", "steps": [{"name": null, "description": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "text": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matched text has a font size of at least 30pt.", "text": "The matched text has a font size of at least 30pt.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Car content slides exists and have correct cars.", "description": "Car content slides exists and have correct cars.", "steps": [{"name": null, "description": "The browsing history contains a visit to the article talking about 2005 best coupes.", "text": "The browsing history contains a visit to the article talking about 2005 best coupes.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "text": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "category": "structural_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Car slides meet the requirements.", "description": "Car slides meet the requirements.", "steps": [{"name": null, "description": "The browsing history contains a visit to the corresponding KBB vehicle page.", "text": "The browsing history contains a visit to the corresponding KBB vehicle page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The make and model of the car are listed in the slide title.", "text": "The make and model of the car are listed in the slide title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A picture of the correct model is found.", "text": "A picture of the correct model is found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The car picture takes up at least 50% of the slide.", "text": "The car picture takes up at least 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "text": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Sticker price matches the price listed on Kelly Blue Book.", "text": "Sticker price matches the price listed on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "text": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Horsepower stat matches the listed value on Kelly Blue Book.", "text": "Horsepower stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a user review platform is provided.", "text": "A URL to a user review platform is provided.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the user review platform URL.", "text": "The browsing history contains a visit to the user review platform URL.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "User average rating matches the listed value on the user review platform.", "text": "User average rating matches the listed value on the user review platform.", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 4, "title": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "description": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "steps": [{"name": null, "description": "The slide title denotes that the slide contains the best car stats.", "text": "The slide title denotes that the slide contains the best car stats.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The correct lowest-price car is listed.", "text": "The correct lowest-price car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest MPG car is listed.", "text": "The correct highest MPG car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest horsepower car is listed.", "text": "The correct highest horsepower car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The most highly-rated car is listed.", "text": "The most highly-rated car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_4/checkpoints.md"}
{"task_id": "knows.slides_29_buy_car_pres.5", "short_id": "29e", "family": "slides_29_buy_car_pres", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Personal", "domain": "Purchase Decision", "prompt": "Create a slides presentation comparing different cars that can be categorized as a sports car. First, find an online article talking about at least 5 different 2010 models. Make a title slide called \"Comparing Different Cool Cars to Buy\" with font size atleast 30pt. Then for each of at least 5 cars of that category, make a slide for each. Start with the make and model as the slide title. Then, for each car, go to Kelly Blue Book and find the sticker price, fuel efficiency and horsepower. Then, go to Kelly Blue Book for user reviews and get the average rating for each. Include the link to the review website in the slide. For each car, add a picture of the car that takes up at least 50% of the slide. The picture must be from the article or Kelly Blue Book. Finally, on the last slide, list which car had the best stats in each category (sticker price, fuel efficiency, horsepower, user rating).", "checkpoints": [{"n": 1, "title": "Title slide contains the correct text.", "description": "Title slide contains the correct text.", "steps": [{"name": null, "description": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "text": "Exact match on \"Comparing Different Cool Cars to Buy\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The matched text has a font size of at least 30pt.", "text": "The matched text has a font size of at least 30pt.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Car content slides exists and have correct cars.", "description": "Car content slides exists and have correct cars.", "steps": [{"name": null, "description": "The browsing history contains a visit to the article talking about 2010 best sports cars.", "text": "The browsing history contains a visit to the article talking about 2010 best sports cars.", "category": "web_visit_check", "llm_vlm_primary": true}, {"name": null, "description": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "text": "There are at least 5 slides, each contains the name of a car that meets the category and year. (5pts total).", "category": "structural_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Car slides meet the requirements.", "description": "Car slides meet the requirements.", "steps": [{"name": null, "description": "The browsing history contains a visit to the corresponding KBB vehicle page.", "text": "The browsing history contains a visit to the corresponding KBB vehicle page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The make and model of the car are listed in the slide title.", "text": "The make and model of the car are listed in the slide title.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A picture of the correct model is found.", "text": "A picture of the correct model is found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The car picture takes up at least 50% of the slide.", "text": "The car picture takes up at least 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "text": "The car picture does not overlap the title text or the stats/URL combined text box (>20% overlap with rendered text region fails).", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Sticker price matches the price listed on Kelly Blue Book.", "text": "Sticker price matches the price listed on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "text": "Fuel efficiency stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Horsepower stat matches the listed value on Kelly Blue Book.", "text": "Horsepower stat matches the listed value on Kelly Blue Book.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a user review platform is provided.", "text": "A URL to a user review platform is provided.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the user review platform URL.", "text": "The browsing history contains a visit to the user review platform URL.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "User average rating matches the listed value on the user review platform.", "text": "User average rating matches the listed value on the user review platform.", "category": "information_retrieval", "llm_vlm_primary": true}]}, {"n": 4, "title": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "description": "The last slide meets the requirements i.e. lists which car had the best stats in each category.", "steps": [{"name": null, "description": "The slide title denotes that the slide contains the best car stats.", "text": "The slide title denotes that the slide contains the best car stats.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The correct lowest-price car is listed.", "text": "The correct lowest-price car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest MPG car is listed.", "text": "The correct highest MPG car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The correct highest horsepower car is listed.", "text": "The correct highest horsepower car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The most highly-rated car is listed.", "text": "The most highly-rated car is listed.", "category": "information_retrieval", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 20, "total_points": 68, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_29_buy_car_pres/instance_5/checkpoints.md"}
{"task_id": "knows.slides_30_wikipedia_photos.1", "short_id": "30a", "family": "slides_30_Work_Wikipedia_Photos", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Work", "domain": "Photo Directory", "prompt": "I represent several clients in Science and want to help replace their bad Wikipedia pages. I need to make a presentation on this. Could you please look at my client list at https://docs.google.com/document/d/1-hzVzV3ZfIgaIa2GDdCHp-X0BOmcklrD8Y2AaaUe5aA/edit?usp=sharing and make a presentation for me? The title should be \"Why we need new Wikipedia headshots\" and have a picture on the top right taking the majority of the slide of Tom Hanks from their wikipedia page. Then, for each of my clients, make a slide with their name at the top, their current wikipedia picture on the left. Then, find images on the internet of each client and put it on the right side so it is symmetric with their wikipedia picture. Finally, on the last slide, copy all of the wikipedia urls of the client pages and then the webpage of a photographer in San Francisco.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"Why we need new Wikipedia headshots\" found.", "text": "Exact match on \"Why we need new Wikipedia headshots\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The wiki image of Tom Hanks is present on the slide.", "text": "The wiki image of Tom Hanks is present on the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Tom Hanks image is positioned in the top right area of the slide.", "text": "The Tom Hanks image is positioned in the top right area of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The Tom Hanks image takes up more than 50% of the slide.", "text": "The Tom Hanks image takes up more than 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Client slides meet the requirements.", "description": "Client slides meet the requirements.", "steps": [{"name": null, "description": "The client's name is found at the top of the slide.", "text": "The client's name is found at the top of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "An image from the client's Wikipedia page is found on the left side of the slide.", "text": "An image from the client's Wikipedia page is found on the left side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "A different image of the client is found on the right side of the slide.", "text": "A different image of the client is found on the right side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "text": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "The last slide contains all required URLs.", "description": "The last slide contains all required URLs.", "steps": [{"name": null, "description": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "text": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a photographer's webpage is present.", "text": "A URL to a photographer's webpage is present.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The photographer is based in San Francisco.", "text": "The photographer is based in San Francisco.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Browsing history shows required page visits.", "description": "Browsing history shows required page visits.", "steps": [{"name": null, "description": "The browsing history contains a visit to Tom Hanks' Wikipedia page.", "text": "The browsing history contains a visit to Tom Hanks' Wikipedia page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the Google Doc containing the client list.", "text": "The browsing history contains a visit to the Google Doc containing the client list.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "text": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 14, "total_points": 62, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/document/d/1-hzVzV3ZfIgaIa2GDdCHp-X0BOmcklrD8Y2AaaUe5aA/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_1/checkpoints.md"}
{"task_id": "knows.slides_30_wikipedia_photos.2", "short_id": "30b", "family": "slides_30_Work_Wikipedia_Photos", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Work", "domain": "Photo Directory", "prompt": "I represent several clients in Music and want to help replace their bad Wikipedia pages. I need to make a presentation on this. Could you please look at my client list at https://docs.google.com/document/d/1JhW_QPaUUHEWMn8vn_Z7PIin4omzTfif6JqujZV0wvs/edit?usp=sharing and make a presentation for me? The title should be \"Why we need new Wikipedia headshots\" and have a picture on the top right taking the majority of the slide of Madonna from her wikipedia page. Then, for each of my clients, make a slide with their name at the top, their current wikipedia picture on the left. Then, find images on the internet of each client and put it on the right side so it is symmetric with their wikipedia picture. Finally, on the last slide, copy all of the wikipedia urls of the client pages and then the webpage of a photographer in Springfield, MA.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"Why we need new Wikipedia headshots\" found.", "text": "Exact match on \"Why we need new Wikipedia headshots\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The wiki image of Madonna is present on the slide.", "text": "The wiki image of Madonna is present on the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Madonna image is positioned in the top right area of the slide.", "text": "The Madonna image is positioned in the top right area of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The Madonna image takes up more than 50% of the slide.", "text": "The Madonna image takes up more than 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Client slides meet the requirements.", "description": "Client slides meet the requirements.", "steps": [{"name": null, "description": "The client's name is found at the top of the slide.", "text": "The client's name is found at the top of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "An image from the client's Wikipedia page is found on the left side of the slide.", "text": "An image from the client's Wikipedia page is found on the left side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "A different image of the client is found on the right side of the slide.", "text": "A different image of the client is found on the right side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "text": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "The last slide contains all required URLs.", "description": "The last slide contains all required URLs.", "steps": [{"name": null, "description": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "text": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a photographer's webpage is present.", "text": "A URL to a photographer's webpage is present.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The photographer is based in Springfield, MA.", "text": "The photographer is based in Springfield, MA.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Browsing history shows required page visits.", "description": "Browsing history shows required page visits.", "steps": [{"name": null, "description": "The browsing history contains a visit to Madonna's Wikipedia page.", "text": "The browsing history contains a visit to Madonna's Wikipedia page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the Google Doc containing the client list.", "text": "The browsing history contains a visit to the Google Doc containing the client list.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "text": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 14, "total_points": 62, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/document/d/1JhW_QPaUUHEWMn8vn_Z7PIin4omzTfif6JqujZV0wvs/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_2/checkpoints.md"}
{"task_id": "knows.slides_30_wikipedia_photos.3", "short_id": "30c", "family": "slides_30_Work_Wikipedia_Photos", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Work", "domain": "Photo Directory", "prompt": "I represent several clients in K-Pop and want to help replace their bad Wikipedia pages. I need to make a presentation on this. Could you please look at my client list at https://docs.google.com/document/d/1dI04kbbBNHxEOsfeYjAdD7AnNYhcnaFVN2Vc-ZcE10w/edit?usp=sharing and make a presentation for me? The title should be \"Why we need new Wikipedia headshots\" and have a picture on the top right taking the majority of the slide of Psy from his wikipedia page. Then, for each of my clients, make a slide with their name at the top, their current wikipedia picture on the left. Then, find images on the internet of each client and put it on the right side so it is symmetric with their wikipedia picture. Finally, on the last slide, copy all of the wikipedia urls of the client pages and then the webpage of a photographer in Seoul.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"Why we need new Wikipedia headshots\" found.", "text": "Exact match on \"Why we need new Wikipedia headshots\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The wiki image of Psy is present on the slide.", "text": "The wiki image of Psy is present on the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Psy image is positioned in the top right area of the slide.", "text": "The Psy image is positioned in the top right area of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The Psy image takes up more than 50% of the slide.", "text": "The Psy image takes up more than 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Client slides meet the requirements.", "description": "Client slides meet the requirements.", "steps": [{"name": null, "description": "The client's name is found at the top of the slide.", "text": "The client's name is found at the top of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "An image from the client's Wikipedia page is found on the left side of the slide.", "text": "An image from the client's Wikipedia page is found on the left side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "A different image of the client is found on the right side of the slide.", "text": "A different image of the client is found on the right side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "text": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "The last slide contains all required URLs.", "description": "The last slide contains all required URLs.", "steps": [{"name": null, "description": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "text": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a photographer's webpage is present.", "text": "A URL to a photographer's webpage is present.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The photographer is based in Seoul.", "text": "The photographer is based in Seoul.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Browsing history shows required page visits.", "description": "Browsing history shows required page visits.", "steps": [{"name": null, "description": "The browsing history contains a visit to Psy's Wikipedia page.", "text": "The browsing history contains a visit to Psy's Wikipedia page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the Google Doc containing the client list.", "text": "The browsing history contains a visit to the Google Doc containing the client list.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "text": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 14, "total_points": 62, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/document/d/1dI04kbbBNHxEOsfeYjAdD7AnNYhcnaFVN2Vc-ZcE10w/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_3/checkpoints.md"}
{"task_id": "knows.slides_30_wikipedia_photos.4", "short_id": "30d", "family": "slides_30_Work_Wikipedia_Photos", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Work", "domain": "Photo Directory", "prompt": "I represent several clients in Fashion and want to help replace their bad Wikipedia pages. I need to make a presentation on this. Could you please look at my client list at https://docs.google.com/document/d/1DEeIchM-qtTYFXrLxwpnBmiGQEnFaixNFYag3qiFCp4/edit?usp=sharing and make a presentation for me? The title should be \"Why we need new Wikipedia headshots\" and have a picture on the top right taking the majority of the slide of Anna Wintour from her wikipedia page. Then, for each of my clients, make a slide with their name at the top, their current wikipedia picture on the left. Then, find images on the internet of each client and put it on the right side so it is symmetric with their wikipedia picture. Finally, on the last slide, copy all of the wikipedia urls of the client pages and then the webpage of a photographer in Milan.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"Why we need new Wikipedia headshots\" found.", "text": "Exact match on \"Why we need new Wikipedia headshots\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The wiki image of Anna Wintour is present on the slide.", "text": "The wiki image of Anna Wintour is present on the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Anna Wintour image is positioned in the top right area of the slide.", "text": "The Anna Wintour image is positioned in the top right area of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The Anna Wintour image takes up more than 50% of the slide.", "text": "The Anna Wintour image takes up more than 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Client slides meet the requirements.", "description": "Client slides meet the requirements.", "steps": [{"name": null, "description": "The client's name is found at the top of the slide.", "text": "The client's name is found at the top of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "An image from the client's Wikipedia page is found on the left side of the slide.", "text": "An image from the client's Wikipedia page is found on the left side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "A different image of the client is found on the right side of the slide.", "text": "A different image of the client is found on the right side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "text": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "The last slide contains all required URLs.", "description": "The last slide contains all required URLs.", "steps": [{"name": null, "description": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "text": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a photographer's webpage is present.", "text": "A URL to a photographer's webpage is present.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The photographer is based in Milan.", "text": "The photographer is based in Milan.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Browsing history shows required page visits.", "description": "Browsing history shows required page visits.", "steps": [{"name": null, "description": "The browsing history contains a visit to Anna Wintour's Wikipedia page.", "text": "The browsing history contains a visit to Anna Wintour's Wikipedia page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the Google Doc containing the client list.", "text": "The browsing history contains a visit to the Google Doc containing the client list.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "text": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 14, "total_points": 62, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/document/d/1DEeIchM-qtTYFXrLxwpnBmiGQEnFaixNFYag3qiFCp4/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_4/checkpoints.md"}
{"task_id": "knows.slides_30_wikipedia_photos.5", "short_id": "30e", "family": "slides_30_Work_Wikipedia_Photos", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Work", "domain": "Photo Directory", "prompt": "I represent several clients in Dance and want to help replace their bad Wikipedia pages. I need to make a presentation on this. Could you please look at my client list at https://docs.google.com/document/d/1M7RH5WK4Y61pQ_2V6dJtlcOSn37I2HXRMqRITJSj-9A/edit?usp=sharing and make a presentation for me? The title should be \"Why we need new Wikipedia headshots\" and have a picture on the top right taking the majority of the slide of Anne Teresa De Keersmaeker from her wikipedia page. Then, for each of my clients, make a slide with their name at the top, their current wikipedia picture on the left. Then, find images on the internet of each client and put it on the right side so it is symmetric with their wikipedia picture. Finally, on the last slide, copy all of the wikipedia urls of the client pages and then the webpage of a photographer in Tel Aviv.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"Why we need new Wikipedia headshots\" found.", "text": "Exact match on \"Why we need new Wikipedia headshots\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "The wiki image of Anne Teresa De Keersmaeker is present on the slide.", "text": "The wiki image of Anne Teresa De Keersmaeker is present on the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Anne Teresa De Keersmaeker image is positioned in the top right area of the slide.", "text": "The Anne Teresa De Keersmaeker image is positioned in the top right area of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The Anne Teresa De Keersmaeker image takes up more than 50% of the slide.", "text": "The Anne Teresa De Keersmaeker image takes up more than 50% of the slide.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Client slides meet the requirements.", "description": "Client slides meet the requirements.", "steps": [{"name": null, "description": "The client's name is found at the top of the slide.", "text": "The client's name is found at the top of the slide.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "An image from the client's Wikipedia page is found on the left side of the slide.", "text": "An image from the client's Wikipedia page is found on the left side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "A different image of the client is found on the right side of the slide.", "text": "A different image of the client is found on the right side of the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "text": "The Wikipedia image and the internet image are approximately symmetric in vertical position.", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "The last slide contains all required URLs.", "description": "The last slide contains all required URLs.", "steps": [{"name": null, "description": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "text": "A Wikipedia URL is present for each client from the Google Doc. (9 pts total)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "A URL to a photographer's webpage is present.", "text": "A URL to a photographer's webpage is present.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "The photographer is based in Tel Aviv.", "text": "The photographer is based in Tel Aviv.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Browsing history shows required page visits.", "description": "Browsing history shows required page visits.", "steps": [{"name": null, "description": "The browsing history contains a visit to Anne Teresa De Keersmaeker's Wikipedia page.", "text": "The browsing history contains a visit to Anne Teresa De Keersmaeker's Wikipedia page.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to the Google Doc containing the client list.", "text": "The browsing history contains a visit to the Google Doc containing the client list.", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "text": "The browsing history contains a visit to each client's Wikipedia page (9 pts total).", "category": "web_visit_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 14, "total_points": 62, "requires_browsing_history": true, "source_assets": ["https://docs.google.com/document/d/1M7RH5WK4Y61pQ_2V6dJtlcOSn37I2HXRMqRITJSj-9A/edit?usp=sharing"], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_30_Work_Wikipedia_Photos/instance_5/checkpoints.md"}
{"task_id": "knows.slides_39_personal_lookbook_paintcolors.1", "short_id": "39a", "family": "slides_39_Personal_Lookbook_PaintColors", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Personal", "domain": "Lookbook", "prompt": "Create a Google slides presentation to help us choose colors for our garage. First, on the title slide, find an image of a garage by searching \"garage\" and choosing one of those images. Make Garage the title text of the title slide and make the image as big as possible. Then, for the next 5-10 slides, choose a color. Make the color the title of each of those slides. Then for each slide, search for images of that color and garage and choose images with that color. Put two of these images on the slide, one on the bottom left and one on the bottom right. Ensure that each image's source URL is present in the image's ALT text. Finally, on the final slide, choose whichever color from the previous slides you like best. Then on that final slide write in big font \"COLOR is the best choice\" and make the background color of only that slide that same color.", "checkpoints": [{"n": 1, "title": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "description": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "steps": [{"name": null, "description": "Verify that the room/project is put as the title (2 pts)", "text": "Verify that the room/project is put as the title (2 pts)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that exactly one image was placed on the title slide (2 pt)", "text": "Confirm that exactly one image was placed on the title slide (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the image covers at least 70% of the slide area (4 pt)", "text": "Check that the image covers at least 70% of the slide area (4 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "text": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "text": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "description": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "steps": [{"name": null, "description": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "text": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify each content slide has a color name as its title (10 pt, proportional)", "text": "Verify each content slide has a color name as its title (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "text": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "description": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "steps": [{"name": null, "description": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "text": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that exactly two images appear on each color slide (10 pt, proportional)", "text": "Check that exactly two images appear on each color slide (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "text": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "text": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "text": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "text": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the two images for a color are unique (10 pt, proportional)", "text": "Check that the two images for a color are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "description": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "steps": [{"name": null, "description": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "text": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "text": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "text": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "text": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 122, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_1/checkpoints.md"}
{"task_id": "knows.slides_39_personal_lookbook_paintcolors.2", "short_id": "39b", "family": "slides_39_Personal_Lookbook_PaintColors", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Personal", "domain": "Lookbook", "prompt": "Create a Google slides presentation to help us choose colors for our kitchen. First, on the title slide, find an image of a kitchen by searching \"kitchen\" and choosing one of those images. Make Kitchen the title text of the title slide and make the image as big as possible. Then, for the next 5-10 slides, choose a color. Make the color the title of each of those slides. Then for each slide, search for images of that color and kitchen and choose images with that color. Put two of these images on the slide, one on the bottom left and one on the bottom right. Ensure that each image's source URL is present in the image's ALT text. Finally, on the final slide, choose whichever color from the previous slides you like best. Then on that final slide write in big font \"COLOR is the best choice\" and make the background color of only that slide that same color.", "checkpoints": [{"n": 1, "title": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "description": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "steps": [{"name": null, "description": "Verify that the room/project is put as the title (2 pts)", "text": "Verify that the room/project is put as the title (2 pts)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that exactly one image was placed on the title slide (2 pt)", "text": "Confirm that exactly one image was placed on the title slide (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the image covers at least 70% of the slide area (4 pt)", "text": "Check that the image covers at least 70% of the slide area (4 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "text": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "text": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "description": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "steps": [{"name": null, "description": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "text": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify each content slide has a color name as its title (10 pt, proportional)", "text": "Verify each content slide has a color name as its title (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "text": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "description": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "steps": [{"name": null, "description": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "text": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that exactly two images appear on each color slide (10 pt, proportional)", "text": "Check that exactly two images appear on each color slide (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "text": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "text": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "text": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "text": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the two images for a color are unique (10 pt, proportional)", "text": "Check that the two images for a color are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "description": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "steps": [{"name": null, "description": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "text": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "text": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "text": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "text": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 122, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_2/checkpoints.md"}
{"task_id": "knows.slides_39_personal_lookbook_paintcolors.3", "short_id": "39c", "family": "slides_39_Personal_Lookbook_PaintColors", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Personal", "domain": "Lookbook", "prompt": "Create a Google slides presentation to help us choose colors for our bedroom. First, on the title slide, find an image of a bedroom by searching \"bedroom\" and choosing one of those images. Make Bedroom the title text of the title slide and make the image as big as possible. Then, for the next 5-10 slides, choose a color. Make the color the title of each of those slides. Then for each slide, search for images of that color and bedroom and choose images with that color. Put two of these images on the slide, one on the bottom left and one on the bottom right. Ensure that each image's source URL is present in the image's ALT text. Finally, on the final slide, choose whichever color from the previous slides you like best. Then on that final slide write in big font \"COLOR is the best choice\" and make the background color of only that slide that same color.", "checkpoints": [{"n": 1, "title": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "description": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "steps": [{"name": null, "description": "Verify that the room/project is put as the title (2 pts)", "text": "Verify that the room/project is put as the title (2 pts)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that exactly one image was placed on the title slide (2 pt)", "text": "Confirm that exactly one image was placed on the title slide (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the image covers at least 70% of the slide area (4 pt)", "text": "Check that the image covers at least 70% of the slide area (4 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "text": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "text": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "description": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "steps": [{"name": null, "description": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "text": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify each content slide has a color name as its title (10 pt, proportional)", "text": "Verify each content slide has a color name as its title (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "text": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "description": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "steps": [{"name": null, "description": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "text": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that exactly two images appear on each color slide (10 pt, proportional)", "text": "Check that exactly two images appear on each color slide (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "text": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "text": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "text": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "text": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the two images for a color are unique (10 pt, proportional)", "text": "Check that the two images for a color are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "description": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "steps": [{"name": null, "description": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "text": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "text": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "text": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "text": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 122, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_3/checkpoints.md"}
{"task_id": "knows.slides_39_personal_lookbook_paintcolors.4", "short_id": "39d", "family": "slides_39_Personal_Lookbook_PaintColors", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Personal", "domain": "Lookbook", "prompt": "Create a Google slides presentation to help us choose colors for our living room. First, on the title slide, find an image of a living room by searching \"living room\" and choosing one of those images. Make Living Room the title text of the title slide and make the image as big as possible. Then, for the next 5-10 slides, choose a color. Make the color the title of each of those slides. Then for each slide, search for images of that color and living room and choose images with that color. Put two of these images on the slide, one on the bottom left and one on the bottom right. Ensure that each image's source URL is present in the image's ALT text. Finally, on the final slide, choose whichever color from the previous slides you like best. Then on that final slide write in big font \"COLOR is the best choice\" and make the background color of only that slide that same color.", "checkpoints": [{"n": 1, "title": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "description": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "steps": [{"name": null, "description": "Verify that the room/project is put as the title (2 pts)", "text": "Verify that the room/project is put as the title (2 pts)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that exactly one image was placed on the title slide (2 pt)", "text": "Confirm that exactly one image was placed on the title slide (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the image covers at least 70% of the slide area (4 pt)", "text": "Check that the image covers at least 70% of the slide area (4 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "text": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "text": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "description": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "steps": [{"name": null, "description": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "text": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify each content slide has a color name as its title (10 pt, proportional)", "text": "Verify each content slide has a color name as its title (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "text": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "description": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "steps": [{"name": null, "description": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "text": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that exactly two images appear on each color slide (10 pt, proportional)", "text": "Check that exactly two images appear on each color slide (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "text": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "text": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "text": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "text": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the two images for a color are unique (10 pt, proportional)", "text": "Check that the two images for a color are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "description": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "steps": [{"name": null, "description": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "text": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "text": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "text": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "text": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 122, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_4/checkpoints.md"}
{"task_id": "knows.slides_39_personal_lookbook_paintcolors.5", "short_id": "39e", "family": "slides_39_Personal_Lookbook_PaintColors", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Personal", "domain": "Lookbook", "prompt": "Create a Google slides presentation to help us choose colors for our home office. First, on the title slide, find an image of a home office by searching \"home office\" and choosing one of those images. Make home office the title text of the title slide and make the image as big as possible. Then, for the next 5-10 slides, choose a color. Make the color the title of each of those slides. Then for each slide, search for images of that color and home office and choose images with that color. Put two of these images on the slide, one on the bottom left and one on the bottom right. Ensure that each image's source URL is present in the image's ALT text. Finally, on the final slide, choose whichever color from the previous slides you like best. Then on that final slide write in big font \"COLOR is the best choice\" and make the background color of only that slide that same color.", "checkpoints": [{"n": 1, "title": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "description": "Check that the title slide contains an appropriate high-quality image of a room/project that fills most of the slide space.", "steps": [{"name": null, "description": "Verify that the room/project is put as the title (2 pts)", "text": "Verify that the room/project is put as the title (2 pts)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that exactly one image was placed on the title slide (2 pt)", "text": "Confirm that exactly one image was placed on the title slide (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the image covers at least 70% of the slide area (4 pt)", "text": "Check that the image covers at least 70% of the slide area (4 pt)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "text": "Verify that the agent searched for the room/project type in browsing history (2 pt)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "text": "Verify image relevance to the room/project type (VLM judge) (2 pt)", "category": "visual_check", "llm_vlm_primary": true}]}, {"n": 2, "title": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "description": "Check that 5-10 color selection slides were created with appropriate color names as titles.", "steps": [{"name": null, "description": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "text": "Count the number of color slides to ensure it falls within the 5-10 range (10 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify each content slide has a color name as its title (10 pt, proportional)", "text": "Verify each content slide has a color name as its title (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "text": "Confirm that the color names are distinct and appropriate for interior design of the room/project (LLM judge) (10 pt, proportional)", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "description": "Verify that each color slide contains two relevant images positioned correctly on the slide, with proper source attribution.", "steps": [{"name": null, "description": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "text": "Confirm that the agent searched for images combining the specific color with the room/project type in browsing history (10 pt, proportional)", "category": "web_visit_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that exactly two images appear on each color slide (10 pt, proportional)", "text": "Check that exactly two images appear on each color slide (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "text": "Verify image positioning: one image in bottom left, one image in bottom right (10 pt, proportional)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "text": "Ensure images are relevant to both the color theme and the room/project type (VLM judge) (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "text": "Verify that each image has a source URL in its ALT text (10 pt, proportional)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "text": "Confirm that the ALT text source URL leads to the same image as displayed on the slide (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the two images for a color are unique (10 pt, proportional)", "text": "Check that the two images for a color are unique (10 pt, proportional)", "category": "visual_check", "llm_vlm_primary": false}]}, {"n": 4, "title": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "description": "Check that a final recommendation slide was created as the last slide with the agent's color choice clearly stated.", "steps": [{"name": null, "description": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "text": "Verify that a recommendation slide exists and is the last slide in the presentation (2 pt)", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "text": "Check that the text \"{COLOR} is the best choice\" appears on the slide (case-insensitive match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "text": "Confirm that the chosen color matches one of the previously presented color slide options (exact match) (3 pt)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "text": "Verify that the recommendation text is displayed in large font (at least 18pt) (2 pt)", "category": "formatting_check", "llm_vlm_primary": false}]}], "n_checkpoints": 4, "n_steps": 19, "total_points": 122, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_39_Personal_Lookbook_PaintColors/instance_5/checkpoints.md"}
{"task_id": "knows.slides_42_product_comparison.1", "short_id": "42a", "family": "slides_42_personal_none_product_comparison", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Personal", "domain": "Product Comparison", "prompt": "My sister, Kathy, has graduated from high school and will be attending the University of Utah in a few months. I want to buy her an laptop as a graduation gift and need help deciding between three options: M4 MacBook Air, Lenovo Yoga 9i Aura Edition, and Surface Laptop.\n\nObjective\nCreate a comprehensive, visually appealing slide deck that compares these three devices specifically for college student use. The presentation should be engaging and supportive rather than dry or overly technical, helping both my sister and me make an informed decision.\n\nTarget Audience\nPrimary: Recent high school graduate heading to college\nSecondary: Family member making the purchase decision\n\nDesign Requirements\nTheme: Use the University of Utah's official colors throughout the presentation\nTone: Exciting, supportive, and easy to understand\nFocus: College student needs and lifestyle\n\nSlide Structure and Content\nSlide 1: Title Slide\n\nTitle (bold): \"A Gift for Kathy!\"\nSubtitle: \"A Comparison of the M4 MacBook Air, Lenovo Yoga 9i Aura Edition, and Surface Laptop\"\nImage: Photo representing the University of Utah (campus scene)\nThe title and subtitle are to the left, and the image is to the right of the slide\n\nSlide 2: The Challenge & The Goal\n\nExplain the decision-making challenge\nDefine what we're trying to achieve\n\nSlide 3: How We'll Judge (The Criteria)\n\nList evaluation criteria relevant to college students:\n\nDurability for campus life\nBattery life\nPortability\nPerformance for academic tasks\nBudget considerations\nOther relevant factors\n\nSlides 4-6: Deep Dive on Each Device\n\nOne slide per device ([Option1], [Option2], [Option3])\nContent per slide:\nKey features and specifications\nPros and cons, w.r.t. college use\nTwo product images showing different angles\n\nAll the information should come from the official website of the option. Cite the resources when needed.\n\nSlide 7: Side-by-Side Comparison\n\nFormat: Three-column comparison table, with devices as column headers\nContent: Feature-by-feature comparison using the established criteria\nColor coding system:\nGreen: Best performer in each category\nYellow: Middle performer\nRed: Lowest performer in that category\n\nSlide 8: Who Is Each Device For? (The Recommendation)\n\nSummarize information on each option and match each device to different student types/needs\nProvide clear recommendations based on different priorities\n\nAdditional Guidelines\nKeep technical jargon to a minimum\nUse student-relatable examples and scenarios\nInclude practical considerations for college life\nEnsure all information is current and accurate\nMake the presentation visually engaging with appropriate use of images, colors, and formatting", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"A Gift for Kathy!\" found.", "text": "Exact match on \"A Gift for Kathy!\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title is in bold.", "text": "Title is in bold.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subtitle correctly lists all 3 device options from the gold list.", "text": "Subtitle correctly lists all 3 device options from the gold list.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Image represents the University of Utah found.", "text": "Image represents the University of Utah found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is to the right of the title.", "text": "Image is to the right of the title.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The university's official colors are used.", "text": "The university's official colors are used.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The Challenge and the goal slides meet requirements.", "description": "The Challenge and the goal slides meet requirements.", "steps": [{"name": null, "description": "At least one line in the slide body explains the challenge of the search.", "text": "At least one line in the slide body explains the challenge of the search.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least one line in the slide body explains the goal of the search.", "text": "At least one line in the slide body explains the goal of the search.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "All evaluation criteria are listed.", "description": "All evaluation criteria are listed.", "steps": [{"name": null, "description": "Battery life found.", "text": "Battery life found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Weight found.", "text": "Weight found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Memory capacity found.", "text": "Memory capacity found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Budget considerations found.", "text": "Budget considerations found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Processor.", "text": "Processor.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "The device slides meet the requirements.", "description": "The device slides meet the requirements.", "steps": [{"name": null, "description": "Title of the slide is the device name.", "text": "Title of the slide is the device name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each slide contains at least one source link.", "text": "Each slide contains at least one source link.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Two product images from different angles found.", "text": "Two product images from different angles found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Key features and specificications section found.", "text": "Key features and specificications section found.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Pros are listed.", "text": "Pros are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Cons are listed.", "text": "Cons are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Product images are from the source link(s) in the slide.", "text": "Product images are from the source link(s) in the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Product key features is accurate according to the sources.", "text": "Product key features is accurate according to the sources.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Side-by-side comparison meets the requirements.", "description": "Side-by-side comparison meets the requirements.", "steps": [{"name": null, "description": "Table has exactly 3 columns.", "text": "Table has exactly 3 columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All devices are included as column headers.", "text": "All devices are included as column headers.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "text": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Battery life is covered.", "text": "Battery life is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Battery life.", "text": "An appropriate color applied for each cell under Battery life.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Weight is covered.", "text": "Weight is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Weight.", "text": "An appropriate color applied for each cell under Weight.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Processor is covered.", "text": "Processor is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Processor.", "text": "An appropriate color applied for each cell under Processor.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Budget consideration is covered.", "text": "Budget consideration is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Budget consideration.", "text": "An appropriate color applied for each cell under Budget consideration.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Memory capacity is covered.", "text": "Memory capacity is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Memory capacity.", "text": "An appropriate color applied for each cell under Memory capacity.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Recommendation slide meets all requirements.", "description": "Recommendation slide meets all requirements.", "steps": [{"name": null, "description": "All three devices found in the slide.", "text": "All three devices found in the slide.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Summaries aligns with the comparison data.", "text": "Summaries aligns with the comparison data.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Recommendations based on different student styles are provided.", "text": "Recommendations based on different student styles are provided.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "description": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "steps": [{"name": null, "description": "Exactly 8 slides found.", "text": "Exactly 8 slides found.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All slides follow the required sequence and titles.", "text": "All slides follow the required sequence and titles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tone is exciting, supportive, and easy to understand (matching the task brief).", "text": "Tone is exciting, supportive, and easy to understand (matching the task brief).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Content is tailored to the recent-grad student audience and the family decision-maker.", "text": "Content is tailored to the recent-grad student audience and the family decision-maker.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 41, "total_points": 57, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_1/checkpoints.md"}
{"task_id": "knows.slides_42_product_comparison.2", "short_id": "42b", "family": "slides_42_personal_none_product_comparison", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Personal", "domain": "Product Comparison", "prompt": "My colleague, James, has just been promoted to Senior Account Manager and will be starting his new role at Deloitte next quarter. I want to buy him a tablet as a congratulatory gift and need help deciding between three options: iPad Pro (M4), Samsung Galaxy Tab S10 Ultra, and Microsoft Surface Pro 11.\nObjective Create a comprehensive, visually appealing slide deck that compares these three devices specifically for a corporate professional's use. The presentation should be polished and persuasive rather than dry or overly technical, helping both my colleague and me make an informed decision.\nTarget Audience Primary: Newly promoted corporate professional entering a client-facing role Secondary: Colleague making the purchase decision\nDesign Requirements Theme: Use Deloitte's official colors (Deloitte Green and black) throughout the presentation Tone: Professional, confident, and easy to understand Focus: Business professional needs and workplace lifestyle\nSlide Structure and Content\nSlide 1: Title Slide\nTitle (bold): \"A Gift for James!\"\nSubtitle: \"A Comparison of the iPad Pro (M4), Samsung Galaxy Tab S10 Ultra, and Microsoft Surface Pro 11\"\nImage: Photo representing Deloitte (office or branding scene)\nThe title and subtitle are to the left, and the image is to the right of the slide\nSlide 2: The Challenge & The Goal\nExplain the decision-making challenge\nDefine what we're trying to achieve\nSlide 3: How We'll Judge (The Criteria)\nList evaluation criteria relevant to business professionals:\nClient presentation capability\nBattery life for long workdays and travel\nPortability and weight for commuting\nPerformance for productivity and multitasking\nSecurity and enterprise compatibility\nOther relevant factors\nSlides 4-6: Deep Dive on Each Device\nOne slide per device (iPad Pro M4, Samsung Galaxy Tab S10 Ultra, Microsoft Surface Pro 11) Content per slide:\nKey features and specifications\nPros and cons with respect to professional/corporate use\nTwo product images showing different angles\nAll the information should come from the official website of each option. Cite the resources when needed.\nSlide 7: Side-by-Side Comparison\nFormat: Three-column comparison table, with devices as column headers\nContent: Feature-by-feature comparison using the established criteria\nColor coding system:\nGreen: Best performer in each category\nYellow: Middle performer\nRed: Lowest performer in that category\nSlide 8: Who Is Each Device For? (The Recommendation)\nSummarize information on each option and match each device to different professional types/needs\nProvide clear recommendations based on different priorities\nAdditional Guidelines\nKeep technical jargon to a minimum\nUse business-relatable examples and scenarios\nInclude practical considerations for corporate life (travel, client meetings, remote work)\nEnsure all information is current and accurate\nMake the presentation visually engaging with appropriate use of images, colors, and formatting", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"A Gift for James!\" found.", "text": "Exact match on \"A Gift for James!\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title is in bold.", "text": "Title is in bold.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subtitle correctly lists all 3 tablet options from the gold list.", "text": "Subtitle correctly lists all 3 tablet options from the gold list.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Image represents Deloitte, such as an office or branding scene, found.", "text": "Image represents Deloitte, such as an office or branding scene, found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is to the right of the title.", "text": "Image is to the right of the title.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Deloitte's official colors, Deloitte Green and black, are used.", "text": "Deloitte's official colors, Deloitte Green and black, are used.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The Challenge and the goal slides meet requirements.", "description": "The Challenge and the goal slides meet requirements.", "steps": [{"name": null, "description": "At least one line in the slide body explains the challenge of the search.", "text": "At least one line in the slide body explains the challenge of the search.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least one line in the slide body explains the goal of the search.", "text": "At least one line in the slide body explains the goal of the search.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "All evaluation criteria are listed.", "description": "All evaluation criteria are listed.", "steps": [{"name": null, "description": "Client presentation capability found.", "text": "Client presentation capability found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Battery life for long workdays and travel found.", "text": "Battery life for long workdays and travel found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Portability and weight for commuting found.", "text": "Portability and weight for commuting found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Performance for productivity and multitasking found.", "text": "Performance for productivity and multitasking found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Security and enterprise compatibility found.", "text": "Security and enterprise compatibility found.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "The tablet slides meet the requirements.", "description": "The tablet slides meet the requirements.", "steps": [{"name": null, "description": "Title of the slide is the device name.", "text": "Title of the slide is the device name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each slide contains at least one source link.", "text": "Each slide contains at least one source link.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Two product images from different angles found.", "text": "Two product images from different angles found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Key features and specifications section found.", "text": "Key features and specifications section found.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Pros are listed.", "text": "Pros are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Cons are listed.", "text": "Cons are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Product images are from the source link(s) in the slide.", "text": "Product images are from the source link(s) in the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Product key features are accurate according to the sources.", "text": "Product key features are accurate according to the sources.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Side-by-side comparison meets the requirements.", "description": "Side-by-side comparison meets the requirements.", "steps": [{"name": null, "description": "Table has exactly 3 columns.", "text": "Table has exactly 3 columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All devices are included as column headers.", "text": "All devices are included as column headers.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "text": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Client presentation capability is covered.", "text": "Client presentation capability is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Client presentation capability.", "text": "An appropriate color applied for each cell under Client presentation capability.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Battery life for long workdays and travel is covered.", "text": "Battery life for long workdays and travel is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Battery life for long workdays and travel.", "text": "An appropriate color applied for each cell under Battery life for long workdays and travel.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Portability and weight for commuting is covered.", "text": "Portability and weight for commuting is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Portability and weight for commuting.", "text": "An appropriate color applied for each cell under Portability and weight for commuting.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Performance for productivity and multitasking is covered.", "text": "Performance for productivity and multitasking is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Performance for productivity and multitasking.", "text": "An appropriate color applied for each cell under Performance for productivity and multitasking.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Security and enterprise compatibility is covered.", "text": "Security and enterprise compatibility is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Security and enterprise compatibility.", "text": "An appropriate color applied for each cell under Security and enterprise compatibility.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Recommendation slide meets all requirements.", "description": "Recommendation slide meets all requirements.", "steps": [{"name": null, "description": "All three tablet options found in the slide.", "text": "All three tablet options found in the slide.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Summaries align with the comparison data.", "text": "Summaries align with the comparison data.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Recommendations based on different professional types or workplace needs are provided.", "text": "Recommendations based on different professional types or workplace needs are provided.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "description": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "steps": [{"name": null, "description": "Exactly 8 slides found.", "text": "Exactly 8 slides found.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All slides follow the required sequence and titles.", "text": "All slides follow the required sequence and titles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tone is professional, confident, and easy to understand (matching the task brief).", "text": "Tone is professional, confident, and easy to understand (matching the task brief).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Content is tailored to the newly promoted corporate professional audience and the colleague decision-maker.", "text": "Content is tailored to the newly promoted corporate professional audience and the colleague decision-maker.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 41, "total_points": 57, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_2/checkpoints.md"}
{"task_id": "knows.slides_42_product_comparison.3", "short_id": "42c", "family": "slides_42_personal_none_product_comparison", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Personal", "domain": "Product Comparison", "prompt": "My student, Maya, has just been accepted into the PhD program in Biochemistry and will be starting her graduate studies at Johns Hopkins University this fall. I want to buy her a scientific notebook/lab journal as a welcome gift and need help deciding between three options: Leuchtturm1917 Hardcover Notebook, Rhodia Webnotebook, and Moleskine Classic Hard Cover Notebook.\nObjective Create a comprehensive, visually appealing slide deck that compares these three notebooks specifically for a graduate researcher's use. The presentation should be encouraging and practical rather than dry or overly technical, helping both my student and me make an informed decision.\nTarget Audience Primary: Incoming PhD student entering a laboratory research environment Secondary: Academic advisor making the purchase decision\nDesign Requirements Theme: Use Johns Hopkins University's official colors (Heritage Blue and black) throughout the presentation Tone: Warm, encouraging, and easy to understand Focus: Graduate student research needs and academic lifestyle\nSlide Structure and Content\nSlide 1: Title Slide\nTitle (bold): \"A Gift for Maya!\"\nSubtitle: \"A Comparison of the Leuchtturm1917 Hardcover Notebook, Rhodia Webnotebook, and Moleskine Classic Hard Cover Notebook\"\nImage: Photo representing Johns Hopkins University (campus or lab scene)\nThe title and subtitle are to the left, and the image is to the right of the slide\nSlide 2: The Challenge & The Goal\nExplain the decision-making challenge\nDefine what we're trying to achieve\nSlide 3: How We'll Judge (The Criteria)\nList evaluation criteria relevant to graduate researchers:\nPaper quality for ink and gel pen compatibility\nDurability for lab and field environments\nPage layout options (dotted, lined, blank) for notes and sketches\nIndexing and organization features\nSize and portability for carrying between lab, library, and class\nOther relevant factors\nSlides 4-6: Deep Dive on Each Notebook\nOne slide per product (Leuchtturm1917, Rhodia Webnotebook, Moleskine Classic) Content per slide:\nKey features and specifications\nPros and cons with respect to graduate research use\nTwo product images showing different angles\nAll the information should come from the official website of each option. Cite the resources when needed.\nSlide 7: Side-by-Side Comparison\nFormat: Three-column comparison table, with notebooks as column headers\nContent: Feature-by-feature comparison using the established criteria\nColor coding system:\nGreen: Best performer in each category\nYellow: Middle performer\nRed: Lowest performer in that category\nSlide 8: Who Is Each Notebook For? (The Recommendation)\nSummarize information on each option and match each notebook to different researcher types/needs\nProvide clear recommendations based on different priorities\nAdditional Guidelines\nKeep technical jargon to a minimum\nUse researcher-relatable examples and scenarios (lab records, lecture notes, field sketches, literature reviews)\nInclude practical considerations for graduate academic life\nEnsure all information is current and accurate\nMake the presentation visually engaging with appropriate use of images, colors, and formatting.", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"A Gift for Maya!\" found.", "text": "Exact match on \"A Gift for Maya!\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title is in bold.", "text": "Title is in bold.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subtitle correctly lists all 3 notebook options from the gold list.", "text": "Subtitle correctly lists all 3 notebook options from the gold list.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Image represents Johns Hopkins University, such as a campus or lab scene, found.", "text": "Image represents Johns Hopkins University, such as a campus or lab scene, found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is to the right of the title.", "text": "Image is to the right of the title.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Johns Hopkins University's official colors, Heritage Blue and black, are used.", "text": "Johns Hopkins University's official colors, Heritage Blue and black, are used.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The Challenge and the goal slides meet requirements.", "description": "The Challenge and the goal slides meet requirements.", "steps": [{"name": null, "description": "At least one line in the slide body explains the challenge of the search.", "text": "At least one line in the slide body explains the challenge of the search.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least one line in the slide body explains the goal of the search.", "text": "At least one line in the slide body explains the goal of the search.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "All evaluation criteria are listed.", "description": "All evaluation criteria are listed.", "steps": [{"name": null, "description": "Paper quality for ink and gel pen compatibility found.", "text": "Paper quality for ink and gel pen compatibility found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Durability for lab and field environments found.", "text": "Durability for lab and field environments found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Page layout options for notes and sketches found.", "text": "Page layout options for notes and sketches found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Indexing and organization features found.", "text": "Indexing and organization features found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Size and portability for carrying between lab, library, and class found.", "text": "Size and portability for carrying between lab, library, and class found.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "The notebook slides meet the requirements.", "description": "The notebook slides meet the requirements.", "steps": [{"name": null, "description": "Title of the slide is the notebook name.", "text": "Title of the slide is the notebook name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each slide contains at least one source link.", "text": "Each slide contains at least one source link.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Two product images from different angles found.", "text": "Two product images from different angles found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Key features and specifications section found.", "text": "Key features and specifications section found.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Pros are listed.", "text": "Pros are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Cons are listed.", "text": "Cons are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Product images are from the source link(s) in the slide.", "text": "Product images are from the source link(s) in the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Product key features are accurate according to the sources.", "text": "Product key features are accurate according to the sources.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Side-by-side comparison meets the requirements.", "description": "Side-by-side comparison meets the requirements.", "steps": [{"name": null, "description": "Table has exactly 3 columns.", "text": "Table has exactly 3 columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All notebooks are included as column headers.", "text": "All notebooks are included as column headers.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "text": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Paper quality for ink and gel pen compatibility is covered.", "text": "Paper quality for ink and gel pen compatibility is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Paper quality for ink and gel pen compatibility.", "text": "An appropriate color applied for each cell under Paper quality for ink and gel pen compatibility.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Durability for lab and field environments is covered.", "text": "Durability for lab and field environments is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Durability for lab and field environments.", "text": "An appropriate color applied for each cell under Durability for lab and field environments.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Page layout options for notes and sketches are covered.", "text": "Page layout options for notes and sketches are covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Page layout options for notes and sketches.", "text": "An appropriate color applied for each cell under Page layout options for notes and sketches.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Indexing and organization features are covered.", "text": "Indexing and organization features are covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Indexing and organization features.", "text": "An appropriate color applied for each cell under Indexing and organization features.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Size and portability for carrying between lab, library, and class is covered.", "text": "Size and portability for carrying between lab, library, and class is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Size and portability for carrying between lab, library, and class.", "text": "An appropriate color applied for each cell under Size and portability for carrying between lab, library, and class.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Recommendation slide meets all requirements.", "description": "Recommendation slide meets all requirements.", "steps": [{"name": null, "description": "All three notebook options found in the slide.", "text": "All three notebook options found in the slide.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Summaries align with the comparison data.", "text": "Summaries align with the comparison data.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Recommendations based on different researcher types or graduate study needs are provided.", "text": "Recommendations based on different researcher types or graduate study needs are provided.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "description": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "steps": [{"name": null, "description": "Exactly 8 slides found.", "text": "Exactly 8 slides found.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All slides follow the required sequence and titles.", "text": "All slides follow the required sequence and titles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tone is warm, encouraging, and easy to understand (matching the task brief).", "text": "Tone is warm, encouraging, and easy to understand (matching the task brief).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Content is tailored to the incoming PhD student audience and the academic advisor decision-maker.", "text": "Content is tailored to the incoming PhD student audience and the academic advisor decision-maker.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 41, "total_points": 57, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_3/checkpoints.md"}
{"task_id": "knows.slides_42_product_comparison.4", "short_id": "42d", "family": "slides_42_personal_none_product_comparison", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Personal", "domain": "Product Comparison", "prompt": "My apprentice, Hiroshi, has just completed his foundational training and will be beginning his journey as a professional tattoo artist at his own studio in Portland, Oregon this spring. I want to buy him a set of tattoo ink as a congratulatory gift and need help deciding between three options: Intenze Ink, Eternal Ink, and Dynamic Ink.\nObjective Create a comprehensive, visually appealing slide deck that compares these three tattoo ink brands specifically for a professional tattoo artist's use. The presentation should be inspiring and craft-focused rather than dry or overly clinical, helping both my apprentice and me make an informed decision.\nTarget Audience Primary: Emerging professional tattoo artist opening their first studio Secondary: Master tattoo artist mentor making the purchase decision\nDesign Requirements Theme: Use deep charcoal, muted gold, and crimson tones reflecting the aesthetic of Portland's independent tattoo culture Tone: Creative, respectful of the craft, and easy to understand Focus: Professional tattoo artist needs and studio lifestyle\nSlide Structure and Content\nSlide 1: Title Slide\nTitle (bold): \"A Gift for Hiroshi!\"\nSubtitle: \"A Comparison of Intenze Ink, Eternal Ink, and Dynamic Ink\"\nImage: Photo representing Portland's independent tattoo art scene (studio or street art scene)\nThe title and subtitle are to the left, and the image is to the right of the slide\nSlide 2: The Challenge & The Goal\nExplain the decision-making challenge\nDefine what we're trying to achieve\nSlide 3: How We'll Judge (The Criteria)\nList evaluation criteria relevant to professional tattoo artists:\nColor vibrancy and consistency across skin tones\nInk longevity and resistance to fading over time\nSafety certifications and vegan/cruelty-free status\nViscosity and needle compatibility for different styles (fine line, traditional, realism)\nRange of available shades and mixing flexibility\nOther relevant factors\nSlides 4-6: Deep Dive on Each Ink Brand\nOne slide per product (Intenze Ink, Eternal Ink, Dynamic Ink) Content per slide:\nKey features and specifications\nPros and cons with respect to professional studio use\nTwo product images showing different angles or product lines\nAll the information should come from the official website of each option. Cite the resources when needed.\nSlide 7: Side-by-Side Comparison\nFormat: Three-column comparison table, with ink brands as column headers\nContent: Feature-by-feature comparison using the established criteria\nColor coding system:\nGreen: Best performer in each category\nYellow: Middle performer\nRed: Lowest performer in that category\nSlide 8: Who Is Each Ink Brand For? (The Recommendation)\nSummarize information on each option and match each ink brand to different artist styles and studio needs\nProvide clear recommendations based on different priorities (e.g., fine line specialists, traditional artists, mixed-style studios)\nAdditional Guidelines\nKeep technical jargon to a minimum\nUse artist-relatable examples and scenarios (client consultations, long sessions, diverse skin tones, style specializations)\nInclude practical considerations for running an independent tattoo studio\nEnsure all information is current and accurate\nMake the presentation visually engaging with appropriate use of images, colors, and formatting", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"A Gift for Hiroshi!\" found.", "text": "Exact match on \"A Gift for Hiroshi!\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title is in bold.", "text": "Title is in bold.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subtitle correctly lists all 3 tattoo ink brand options from the gold list.", "text": "Subtitle correctly lists all 3 tattoo ink brand options from the gold list.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Image represents Portland's independent tattoo art scene, such as a studio or street art scene, found.", "text": "Image represents Portland's independent tattoo art scene, such as a studio or street art scene, found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is to the right of the title.", "text": "Image is to the right of the title.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Deep charcoal, muted gold, and crimson tones are used.", "text": "Deep charcoal, muted gold, and crimson tones are used.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The Challenge and the goal slides meet requirements.", "description": "The Challenge and the goal slides meet requirements.", "steps": [{"name": null, "description": "At least one line in the slide body explains the challenge of the search.", "text": "At least one line in the slide body explains the challenge of the search.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least one line in the slide body explains the goal of the search.", "text": "At least one line in the slide body explains the goal of the search.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "All evaluation criteria are listed.", "description": "All evaluation criteria are listed.", "steps": [{"name": null, "description": "Color vibrancy and consistency across skin tones found.", "text": "Color vibrancy and consistency across skin tones found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Ink longevity and resistance to fading over time found.", "text": "Ink longevity and resistance to fading over time found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Safety certifications and vegan/cruelty-free status found.", "text": "Safety certifications and vegan/cruelty-free status found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Viscosity and needle compatibility for different styles found.", "text": "Viscosity and needle compatibility for different styles found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Range of available shades and mixing flexibility found.", "text": "Range of available shades and mixing flexibility found.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "The tattoo ink brand slides meet the requirements.", "description": "The tattoo ink brand slides meet the requirements.", "steps": [{"name": null, "description": "Title of the slide is the ink brand name.", "text": "Title of the slide is the ink brand name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each slide contains at least one source link.", "text": "Each slide contains at least one source link.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Two product images from different angles found.", "text": "Two product images from different angles found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Key features and specifications section found.", "text": "Key features and specifications section found.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Pros are listed.", "text": "Pros are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Cons are listed.", "text": "Cons are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Product images are from the source link(s) in the slide.", "text": "Product images are from the source link(s) in the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Product key features are accurate according to the sources.", "text": "Product key features are accurate according to the sources.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Side-by-side comparison meets the requirements.", "description": "Side-by-side comparison meets the requirements.", "steps": [{"name": null, "description": "Table has exactly 3 columns.", "text": "Table has exactly 3 columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All ink brands are included as column headers.", "text": "All ink brands are included as column headers.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "text": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Color vibrancy and consistency across skin tones is covered.", "text": "Color vibrancy and consistency across skin tones is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Color vibrancy and consistency across skin tones.", "text": "An appropriate color applied for each cell under Color vibrancy and consistency across skin tones.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Ink longevity and resistance to fading over time is covered.", "text": "Ink longevity and resistance to fading over time is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Ink longevity and resistance to fading over time.", "text": "An appropriate color applied for each cell under Ink longevity and resistance to fading over time.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Safety certifications and vegan/cruelty-free status is covered.", "text": "Safety certifications and vegan/cruelty-free status is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Safety certifications and vegan/cruelty-free status.", "text": "An appropriate color applied for each cell under Safety certifications and vegan/cruelty-free status.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Viscosity and needle compatibility for different styles is covered.", "text": "Viscosity and needle compatibility for different styles is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Viscosity and needle compatibility for different styles.", "text": "An appropriate color applied for each cell under Viscosity and needle compatibility for different styles.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Range of available shades and mixing flexibility is covered.", "text": "Range of available shades and mixing flexibility is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Range of available shades and mixing flexibility.", "text": "An appropriate color applied for each cell under Range of available shades and mixing flexibility.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Recommendation slide meets all requirements.", "description": "Recommendation slide meets all requirements.", "steps": [{"name": null, "description": "All three tattoo ink brand options found in the slide.", "text": "All three tattoo ink brand options found in the slide.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Summaries align with the comparison data.", "text": "Summaries align with the comparison data.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Recommendations based on different artist styles or studio needs are provided.", "text": "Recommendations based on different artist styles or studio needs are provided.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "description": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "steps": [{"name": null, "description": "Exactly 8 slides found.", "text": "Exactly 8 slides found.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All slides follow the required sequence and titles.", "text": "All slides follow the required sequence and titles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tone is creative, respectful of the craft, and easy to understand (matching the task brief).", "text": "Tone is creative, respectful of the craft, and easy to understand (matching the task brief).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Content is tailored to the emerging professional tattoo artist audience and the mentor decision-maker.", "text": "Content is tailored to the emerging professional tattoo artist audience and the mentor decision-maker.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 41, "total_points": 57, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_4/checkpoints.md"}
{"task_id": "knows.slides_42_product_comparison.5", "short_id": "42e", "family": "slides_42_personal_none_product_comparison", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Personal", "domain": "Product Comparison", "prompt": "My teammate, Serena, has just earned her spot on the U.S. National Rowing Team and will be beginning her elite training camp at the Olympic Training Center in Colorado Springs this summer. I want to buy her a pair of rowing gloves as a celebratory gift and need help deciding between three options: The Crew Stop Rowing Gloves, NRS Cove Paddling Gloves, and Harbinger Pro Wrist Wrap Gloves.\nObjective Create a comprehensive, visually appealing slide deck that compares these three rowing gloves specifically for an elite competitive rower's use. The presentation should be energetic and performance-driven rather than dry or overly technical, helping both my teammate and me make an informed decision.\nTarget Audience Primary: Elite athlete beginning national-level competitive rowing training Secondary: Teammate making the purchase decision\nDesign Requirements Theme: Use the U.S. Olympic & Paralympic Committee's official colors (red, white, and navy blue) throughout the presentation Tone: Energetic, celebratory, and easy to understand Focus: Elite athlete performance needs and high-intensity training lifestyle\nSlide Structure and Content\nSlide 1: Title Slide\nTitle (bold): \"A Gift for Serena!\"\nSubtitle: \"A Comparison of The Crew Stop Rowing Gloves, NRS Cove Paddling Gloves, and Harbinger Pro Wrist Wrap Gloves\"\nImage: Photo representing the Olympic Training Center in Colorado Springs (facility or rowing scene)\nThe title and subtitle are to the left, and the image is to the right of the slide\nSlide 2: The Challenge & The Goal\nExplain the decision-making challenge\nDefine what we're trying to achieve\nSlide 3: How We'll Judge (The Criteria)\nList evaluation criteria relevant to elite competitive rowers:\nGrip strength and palm protection during high-repetition strokes\nBreathability and moisture management during long training sessions\nWrist support and injury prevention for heavy loading\nDurability under daily elite training conditions\nFit and dexterity for maintaining proper oar technique\nOther relevant factors\nSlides 4-6: Deep Dive on Each Glove\nOne slide per product (The Crew Stop Rowing Gloves, NRS Cove Paddling Gloves, Harbinger Pro Wrist Wrap Gloves) Content per slide:\nKey features and specifications\nPros and cons with respect to elite rowing training use\nTwo product images showing different angles\nAll the information should come from the official website of each option. Cite the resources when needed.\nSlide 7: Side-by-Side Comparison\nFormat: Three-column comparison table, with gloves as column headers\nContent: Feature-by-feature comparison using the established criteria\nColor coding system:\nGreen: Best performer in each category\nYellow: Middle performer\nRed: Lowest performer in that category\nSlide 8: Who Is Each Glove For? (The Recommendation)\nSummarize information on each option and match each glove to different athlete types and training needs\nProvide clear recommendations based on different priorities (e.g., blister-prone rowers, injury-recovery athletes, technique-focused beginners at the elite level)\nAdditional Guidelines\nKeep technical jargon to a minimum\nUse athlete-relatable examples and scenarios (early morning water sessions, ergometer training, multi-session training days, competition preparation)\nInclude practical considerations for life at an elite training camp\nEnsure all information is current and accurate\nMake the presentation visually engaging with appropriate use of images, colors, and formatting", "checkpoints": [{"n": 1, "title": "Title slide has all required elements.", "description": "Title slide has all required elements.", "steps": [{"name": null, "description": "Exact match on \"A Gift for Serena!\" found.", "text": "Exact match on \"A Gift for Serena!\" found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Title is in bold.", "text": "Title is in bold.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subtitle correctly lists all 3 rowing glove options from the gold list.", "text": "Subtitle correctly lists all 3 rowing glove options from the gold list.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Image represents the Olympic Training Center in Colorado Springs, such as a facility or rowing scene, found.", "text": "Image represents the Olympic Training Center in Colorado Springs, such as a facility or rowing scene, found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Image is to the right of the title.", "text": "Image is to the right of the title.", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "The U.S. Olympic & Paralympic Committee's official colors, red, white, and navy blue, are used.", "text": "The U.S. Olympic & Paralympic Committee's official colors, red, white, and navy blue, are used.", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "The Challenge and the goal slides meet requirements.", "description": "The Challenge and the goal slides meet requirements.", "steps": [{"name": null, "description": "At least one line in the slide body explains the challenge of the search.", "text": "At least one line in the slide body explains the challenge of the search.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least one line in the slide body explains the goal of the search.", "text": "At least one line in the slide body explains the goal of the search.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 3, "title": "All evaluation criteria are listed.", "description": "All evaluation criteria are listed.", "steps": [{"name": null, "description": "Grip strength and palm protection during high-repetition strokes found.", "text": "Grip strength and palm protection during high-repetition strokes found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Breathability and moisture management during long training sessions found.", "text": "Breathability and moisture management during long training sessions found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Wrist support and injury prevention for heavy loading found.", "text": "Wrist support and injury prevention for heavy loading found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Durability under daily elite training conditions found.", "text": "Durability under daily elite training conditions found.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Fit and dexterity for maintaining proper oar technique found.", "text": "Fit and dexterity for maintaining proper oar technique found.", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 4, "title": "The rowing glove slides meet the requirements.", "description": "The rowing glove slides meet the requirements.", "steps": [{"name": null, "description": "Title of the slide is the glove name.", "text": "Title of the slide is the glove name.", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Each slide contains at least one source link.", "text": "Each slide contains at least one source link.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Two product images from different angles found.", "text": "Two product images from different angles found.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Key features and specifications section found.", "text": "Key features and specifications section found.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Pros are listed.", "text": "Pros are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Cons are listed.", "text": "Cons are listed.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "Product images are from the source link(s) in the slide.", "text": "Product images are from the source link(s) in the slide.", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Product key features are accurate according to the sources.", "text": "Product key features are accurate according to the sources.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Side-by-side comparison meets the requirements.", "description": "Side-by-side comparison meets the requirements.", "steps": [{"name": null, "description": "Table has exactly 3 columns.", "text": "Table has exactly 3 columns.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All gloves are included as column headers.", "text": "All gloves are included as column headers.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "text": "Three colors red, yellow, and green are used for the coding scheme in the table content.", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Grip strength and palm protection during high-repetition strokes is covered.", "text": "Grip strength and palm protection during high-repetition strokes is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Grip strength and palm protection during high-repetition strokes.", "text": "An appropriate color applied for each cell under Grip strength and palm protection during high-repetition strokes.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Breathability and moisture management during long training sessions is covered.", "text": "Breathability and moisture management during long training sessions is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Breathability and moisture management during long training sessions.", "text": "An appropriate color applied for each cell under Breathability and moisture management during long training sessions.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Wrist support and injury prevention for heavy loading is covered.", "text": "Wrist support and injury prevention for heavy loading is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Wrist support and injury prevention for heavy loading.", "text": "An appropriate color applied for each cell under Wrist support and injury prevention for heavy loading.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Durability under daily elite training conditions is covered.", "text": "Durability under daily elite training conditions is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Durability under daily elite training conditions.", "text": "An appropriate color applied for each cell under Durability under daily elite training conditions.", "category": "formatting_check", "llm_vlm_primary": true}, {"name": null, "description": "Fit and dexterity for maintaining proper oar technique is covered.", "text": "Fit and dexterity for maintaining proper oar technique is covered.", "category": "structural_check", "llm_vlm_primary": true}, {"name": null, "description": "An appropriate color applied for each cell under Fit and dexterity for maintaining proper oar technique.", "text": "An appropriate color applied for each cell under Fit and dexterity for maintaining proper oar technique.", "category": "formatting_check", "llm_vlm_primary": true}]}, {"n": 6, "title": "Recommendation slide meets all requirements.", "description": "Recommendation slide meets all requirements.", "steps": [{"name": null, "description": "All three rowing glove options found in the slide.", "text": "All three rowing glove options found in the slide.", "category": "information_retrieval", "llm_vlm_primary": true}, {"name": null, "description": "Summaries align with the comparison data.", "text": "Summaries align with the comparison data.", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Recommendations based on different athlete types or training needs are provided.", "text": "Recommendations based on different athlete types or training needs are provided.", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "description": "The overall structure follows the specified task requirements and the deck addresses the stated audience and tone.", "steps": [{"name": null, "description": "Exactly 8 slides found.", "text": "Exactly 8 slides found.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "All slides follow the required sequence and titles.", "text": "All slides follow the required sequence and titles.", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Tone is energetic, celebratory, and easy to understand (matching the task brief).", "text": "Tone is energetic, celebratory, and easy to understand (matching the task brief).", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Content is tailored to the elite competitive rower audience and the teammate decision-maker.", "text": "Content is tailored to the elite competitive rower audience and the teammate decision-maker.", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 41, "total_points": 57, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_42_personal_none_product_comparison/instance_5/checkpoints.md"}
{"task_id": "knows.slides_51_event_announcement_poster.1", "short_id": "51a", "family": "slides_51_event_announcement_poster", "instance": "instance_1", "instance_num": 1, "artifact": "Slides", "group": "Personal", "domain": "Party Invite / Event Poster", "prompt": "Create a one-page Google Slide, formatted as a visually appealing event announcement poster.\nThis is an event called Foundations of Web Browser Internals, organized by the University of Utah. It will take place on 07/27/2026 at Warnock Engineering Building Room L101 at the University of Utah. The main theme is how web browsers work, and the featured speaker/performer is Pavel Panchekha.\nRetrieve reliable background information about how web browsers work that is closely related to Pavel Panchekha’s work from a few stable, reputable sources (e.g., universities, professional associations, encyclopedias, major news outlets, and the speaker’s website). Summarize the topic in 1-2 paragraphs that highlight why the event is worth attending and cite the websites you used. Also, introduce the speaker, including their education and current affiliation, and mention 3 of their main professional achievements.\n\nPoster Structure\nHeader (top, centered):\nFoundations of Web Browser Internals, large, bold, in color Black.\n\nSubheader (below header, left-aligned):\n07/27/2026, Warnock Engineering Building Room L101, and hosted University of Utah, styled in a smaller italic font.\n\nLogo (top, right)\nRetrieve the logo of the University of Utah affiliation and insert it to the top right of the poster. Make sure to adjust the logo size so that it’s not too big or too small.\n\nCentral Section (main body, centered):\nSynthesized background on how web browsers work (engaging, persuasive, 1-2 short paragraphs).\n\nRight Sidebar (shaded box, color grey):\n\nSpeaker/performer bio (Pavel Panchekha), including their credentials, achievements, and relevance to how web browsers work.\n\nFooter (bottom, right-aligned):\nContact info of the University of Utah, Pavel Panchekha’s webpage.\n\n\nStyling Requirements\nEnsure visual contrast between header, subheader, and body text.\nUse at least two colors: Black for header elements and Grey for sidebars/highlights.\nKeep consistent spacing and alignment so the document resembles a poster, not just plain text.\nEnsure that citations are placed in speaker notes and say exactly what the are citing in the poster.", "checkpoints": [{"n": 1, "title": "Header & Subheader", "description": "The header and subheader follow proper structure and contain correct information.", "steps": [{"name": null, "description": "Event name text box appears at the top of the page", "text": "Event name text box appears at the top of the page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text is centered on the slide", "text": "Event name text is centered on the slide", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has bold font", "text": "Event name text has bold font", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has font size >= 20 (or whatever it is)", "text": "Event name text has font size >= 20 (or whatever it is)", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text color is black", "text": "Event name text color is black", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name reads \"Foundations of Web Browser Internals\"", "text": "Event name reads \"Foundations of Web Browser Internals\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader exists below the main header", "text": "Subheader exists below the main header", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is left aligned", "text": "Subheader is left aligned", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is in italic font smaller than the main header", "text": "Subheader is in italic font smaller than the main header", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains date (07/27/2026)", "text": "Subheader contains date (07/27/2026)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains location (Warnock Engineering Building Room L101)", "text": "Subheader contains location (Warnock Engineering Building Room L101)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains host (University of Utah)", "text": "Subheader contains host (University of Utah)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Visual contrast is clear between header and subheader text", "text": "Visual contrast is clear between header and subheader text", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Logo", "description": "The logo is properly placed, scaled, and sourced.", "steps": [{"name": null, "description": "The logo is the University of Utah's logo", "text": "The logo is the University of Utah's logo", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is inserted in the top right of the slide", "text": "Logo is inserted in the top right of the slide", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo appropriately scaled (not disproportionately large or small)", "text": "Logo appropriately scaled (not disproportionately large or small)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo does not overlap or crowd header/subheader text", "text": "Logo does not overlap or crowd header/subheader text", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Central Body — Topic Background", "description": "The central body contains an accurate and well-sourced summary of the topic.", "steps": [{"name": null, "description": "Background summary is placed in the main, central body of the poster", "text": "Background summary is placed in the main, central body of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary is 1–2 paragraphs", "text": "Summary is 1–2 paragraphs", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary covers how web browsers work", "text": "Summary covers how web browsers work", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary clearly tied to Pavel Panchekha's work or field of expertise", "text": "Summary clearly tied to Pavel Panchekha's work or field of expertise", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary is engaging and highlights why the event is worth attending", "text": "Summary is engaging and highlights why the event is worth attending", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "text": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Right Sidebar — Speaker Bio", "description": "The sidebar contains correct and verifiable speaker information.", "steps": [{"name": null, "description": "A shaded right-hand sidebar in grey exists", "text": "A shaded right-hand sidebar in grey exists", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Right sidebar contains text", "text": "Right sidebar contains text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speaker's current affiliation (institution/company) is explicitly named and correct according to a reputable source", "text": "Speaker's current affiliation (institution/company) is explicitly named and correct according to a reputable source", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Speaker's educational background (e.g., degree, university) is included and accurate according to a reputable source", "text": "Speaker's educational background (e.g., degree, university) is included and accurate according to a reputable source", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "3 distinct, verifiable professional achievements are listed", "text": "3 distinct, verifiable professional achievements are listed", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sidebar includes connection between the speaker and web browsers", "text": "Sidebar includes connection between the speaker and web browsers", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Footer", "description": "The footer contains correct contact and reference information.", "steps": [{"name": null, "description": "Footer is positioned at the bottom right of the poster", "text": "Footer is positioned at the bottom right of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains University of Utah's contact info", "text": "Footer contains University of Utah's contact info", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains Pavel Panchekha's webpage URL", "text": "Footer contains Pavel Panchekha's webpage URL", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Citations", "description": "All sources are properly cited and traceable.", "steps": [{"name": null, "description": "Citations are present in the speaker notes", "text": "Citations are present in the speaker notes", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sources cited correspond to specific claims in the topic summary", "text": "Sources cited correspond to specific claims in the topic summary", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sources cited correspond to specific claims in the speaker bio", "text": "Sources cited correspond to specific claims in the speaker bio", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "Overall structure", "description": "All boxes have good structure", "steps": [{"name": null, "description": "No text or images are overalapping each other", "text": "No text or images are overalapping each other", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are off of the slide", "text": "No text or images are off of the slide", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 37, "total_points": 185, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_1/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_1/checkpoints.md"}
{"task_id": "knows.slides_51_event_announcement_poster.2", "short_id": "51b", "family": "slides_51_event_announcement_poster", "instance": "instance_2", "instance_num": 2, "artifact": "Slides", "group": "Personal", "domain": "Party Invite / Event Poster", "prompt": "Create a one-page Google Slides formatted as a visually appealing event announcement poster. This is an event called Becoming a National Geographic Photographer, organized by the Center for Creative Photography at the University of Arizona. It will take place on September 15th at 1030 North Olive Road, Tucson. The main theme is Photography, and the featured speaker/performer is Michael Yamashita. Retrieve reliable background information about Photography that is closely related to Michael Yamashita’s work from a few stable, reputable sources (e.g., universities, professional associations, encyclopedias, major news outlets, and the speaker’s website). Summarize the topic in 1-2 paragraphs that highlight why the event is worth attending and cite the websites you used. Also, introduce the speaker, including their education and current affiliation, and mention 5 of their main professional achievements.\n\nPoster Structure:\n\nHeader (top, centered): Becoming a National Geographic Photographer, large, bold, in color Mint Green.\n\nSubheader (below header, left-aligned): September 15th, 1030 North Olive Road, Tucson, and Center for Creative Photography at University of Arizona, styled in a smaller italic font.\n\nLogo (top, right): Retrieve the logo of the Center for Creative Photography at University of Arizona affiliation and insert it to the top right of the poster. Make sure to adjust the logo size so that it’s not too big or too small.\n\nCentral Section (main body, centered): Synthesized background on Photography (engaging, persuasive, 1-2 short paragraphs).\n\nRight Sidebar (shaded box, color Light Orange): Speaker/performer bio (Michael Yamashita), including their credentials, achievements, and relevance to Photography.\n\nFooter (bottom, right-aligned): Contact info of the Center for Creative Photography at University of Arizona, Michael Yamashita’s webpage.\n\n\nStyling Requirements: Ensure visual contrast between header, subheader, and body text. Use at least two colors: Mint Green for header elements and Light Orange for sidebars/highlights. Keep consistent spacing and alignment so the document resembles a poster, not just plain text. Suggest 1-2 short pull quotes, statistics, or facts about Photography that could be highlighted in callout boxes.", "checkpoints": [{"n": 1, "title": "Header & Subheader", "description": "The header and subheader follow proper structure and contain correct information.", "steps": [{"name": null, "description": "Event name text box appears at the top of the page", "text": "Event name text box appears at the top of the page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text is centered on the slide", "text": "Event name text is centered on the slide", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has bold font", "text": "Event name text has bold font", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has font size >= 20", "text": "Event name text has font size >= 20", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text color is **Mint Green**", "text": "Event name text color is **Mint Green**", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name reads \"Becoming a National Geographic Photographer\"", "text": "Event name reads \"Becoming a National Geographic Photographer\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader exists below the main header", "text": "Subheader exists below the main header", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is left-aligned", "text": "Subheader is left-aligned", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is in italic font smaller than the main header", "text": "Subheader is in italic font smaller than the main header", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains date (September 15th)", "text": "Subheader contains date (September 15th)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains location (1030 North Olive Road, Tucson)", "text": "Subheader contains location (1030 North Olive Road, Tucson)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains host (Center for Creative Photography at University of Arizona)", "text": "Subheader contains host (Center for Creative Photography at University of Arizona)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Visual contrast is clear between header and subheader text", "text": "Visual contrast is clear between header and subheader text", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Logo", "description": "The logo is properly placed, scaled, and sourced.", "steps": [{"name": null, "description": "The logo is the **Center for Creative Photography at University of Arizona** logo", "text": "The logo is the **Center for Creative Photography at University of Arizona** logo", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is inserted in the top right of the slide", "text": "Logo is inserted in the top right of the slide", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo appropriately scaled (not disproportionately large or small)", "text": "Logo appropriately scaled (not disproportionately large or small)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo does not overlap or crowd header/subheader text", "text": "Logo does not overlap or crowd header/subheader text", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Central Body — Topic Background", "description": "The central body contains an accurate and well-sourced summary of the topic.", "steps": [{"name": null, "description": "Background summary is placed in the main, central body of the poster", "text": "Background summary is placed in the main, central body of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary is 1–2 paragraphs", "text": "Summary is 1–2 paragraphs", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary covers Photography", "text": "Summary covers Photography", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary clearly tied to Michael Yamashita's work or field of expertise", "text": "Summary clearly tied to Michael Yamashita's work or field of expertise", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary is engaging and highlights why the event is worth attending", "text": "Summary is engaging and highlights why the event is worth attending", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "text": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Right Sidebar — Speaker Bio", "description": "The sidebar contains correct and verifiable speaker information.", "steps": [{"name": null, "description": "A shaded right-hand sidebar in **Light Orange** exists", "text": "A shaded right-hand sidebar in **Light Orange** exists", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Right sidebar contains text", "text": "Right sidebar contains text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speaker's current affiliation (institution/company) is explicitly named and correct according to a reputable source", "text": "Speaker's current affiliation (institution/company) is explicitly named and correct according to a reputable source", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Speaker's educational background (e.g., degree, university) is included and accurate according to a reputable source", "text": "Speaker's educational background (e.g., degree, university) is included and accurate according to a reputable source", "category": "content_check", "llm_vlm_primary": true}, {"name": "5", "description": "distinct, verifiable professional achievements are listed", "text": "**5** distinct, verifiable professional achievements are listed", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sidebar includes connection between the speaker and the field of photography", "text": "Sidebar includes connection between the speaker and the field of photography", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Footer", "description": "The footer contains correct contact and reference information.", "steps": [{"name": null, "description": "Footer is positioned at the bottom right of the poster", "text": "Footer is positioned at the bottom right of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains contact info for the Center for Creative Photography at University of Arizona", "text": "Footer contains contact info for the Center for Creative Photography at University of Arizona", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains Michael Yamashita's webpage URL", "text": "Footer contains Michael Yamashita's webpage URL", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Citations & Callouts", "description": "All sources are properly cited and additional engagement elements are present.", "steps": [{"name": null, "description": "Citations are present in the speaker notes", "text": "Citations are present in the speaker notes", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sources cited correspond to specific claims in the topic summary", "text": "Sources cited correspond to specific claims in the topic summary", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sources cited correspond to specific claims in the speaker bio", "text": "Sources cited correspond to specific claims in the speaker bio", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "1–2 short pull quotes, statistics, or facts about Photography are present in callout boxes", "text": "1–2 short pull quotes, statistics, or facts about Photography are present in callout boxes", "category": "structural_check", "llm_vlm_primary": false}]}, {"n": 7, "title": "Overall Structure", "description": "All boxes have good structure.", "steps": [{"name": null, "description": "No text or images are overlapping each other", "text": "No text or images are overlapping each other", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are off of the slide", "text": "No text or images are off of the slide", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 38, "total_points": 190, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_2/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_2/checkpoints.md"}
{"task_id": "knows.slides_51_event_announcement_poster.3", "short_id": "51c", "family": "slides_51_event_announcement_poster", "instance": "instance_3", "instance_num": 3, "artifact": "Slides", "group": "Personal", "domain": "Party Invite / Event Poster", "prompt": "Create a one-page Google Slides formatted as a visually appealing event announcement poster. This is an event called Diving Into the Deep: A Life of Ocean Exploration, organized by the Scripps Institution of Oceanography at UC San Diego. It will take place on November 8th at 8622 Kennel Way, La Jolla, CA. The main theme is Ocean Conservation, and the featured speaker is Dr. Sylvia Earle.\n\nRetrieve reliable background information about Ocean Conservation that is closely related to Dr. Sylvia Earle's work from stable, reputable sources (e.g., universities, professional associations, encyclopedias, major news outlets, and the speaker's website). Summarize the topic in 1-2 paragraphs that highlight why the event is worth attending and cite the websites you used. Also, introduce the speaker, including their education and current affiliation, and mention 3 of their main professional achievements.\n\nPoster Structure:\n\nHeader (top, centered): Diving Into the Deep: A Life of Ocean Exploration, large, bold, in color Deep Ocean Blue.\n\nSubheader (below header, left-aligned):\nNovember 8th, 8622 Kennel Way, La Jolla, CA, and the Scripps Institution of Oceanography at UC San Diego, styled in a smaller italic font.\n\nLogo (top, right):\nRetrieve the logo of the Scripps Institution of Oceanography at UC San Diego and insert it to the top right of the poster. Make sure to adjust the logo size so that it's not too big or too small.\n\nCentral Section (main body, centered):\nSynthesized background on Ocean Conservation (engaging, persuasive, 1-2 short paragraphs).\n\nRight Sidebar (shaded box, color Seafoam Green):\nSpeaker bio (Dr. Sylvia Earle), including her credentials, achievements, and relevance to Ocean Conservation.\n\nFooter (bottom, right-aligned):\nContact info of the Scripps Institution of Oceanography at UC San Diego, Dr. Sylvia Earle's webpage (missionblue.org).\n\nStyling Requirements: Ensure visual contrast between header, subheader, and body text. Use at least two colors: Deep Ocean Blue for header elements and Seafoam Green for sidebars/highlights. Keep consistent spacing and alignment so the document resembles a poster, not just plain text.", "checkpoints": [{"n": 1, "title": "Header & Subheader", "description": "The header and subheader follow proper structure and contain correct information.", "steps": [{"name": null, "description": "Event name text box appears at the top of the page", "text": "Event name text box appears at the top of the page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text is centered on the slide", "text": "Event name text is centered on the slide", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has bold font", "text": "Event name text has bold font", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has font size >= 20", "text": "Event name text has font size >= 20", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text color is **Deep Ocean Blue**", "text": "Event name text color is **Deep Ocean Blue**", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name reads \"Diving Into the Deep: A Life of Ocean Exploration\"", "text": "Event name reads \"Diving Into the Deep: A Life of Ocean Exploration\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader exists below the main header", "text": "Subheader exists below the main header", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is left-aligned", "text": "Subheader is left-aligned", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is in italic font smaller than the main header", "text": "Subheader is in italic font smaller than the main header", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains date (November 8th)", "text": "Subheader contains date (November 8th)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains location (8622 Kennel Way, La Jolla, CA)", "text": "Subheader contains location (8622 Kennel Way, La Jolla, CA)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains host (Scripps Institution of Oceanography at UC San Diego)", "text": "Subheader contains host (Scripps Institution of Oceanography at UC San Diego)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Visual contrast is clear between header and subheader text", "text": "Visual contrast is clear between header and subheader text", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Logo", "description": "The logo is properly placed, scaled, and sourced.", "steps": [{"name": null, "description": "The logo is the **Scripps Institution of Oceanography at UC San Diego** logo", "text": "The logo is the **Scripps Institution of Oceanography at UC San Diego** logo", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is inserted in the top right of the slide", "text": "Logo is inserted in the top right of the slide", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo appropriately scaled (not disproportionately large or small)", "text": "Logo appropriately scaled (not disproportionately large or small)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo does not overlap or crowd header/subheader text", "text": "Logo does not overlap or crowd header/subheader text", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Central Body — Topic Background", "description": "The central body contains an accurate and well-sourced summary of Ocean Conservation.", "steps": [{"name": null, "description": "Background summary is placed in the main, central body of the poster", "text": "Background summary is placed in the main, central body of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary is 1–2 paragraphs", "text": "Summary is 1–2 paragraphs", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary covers Ocean Conservation", "text": "Summary covers Ocean Conservation", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary clearly tied to Dr. Sylvia Earle's work or field of expertise (e.g., Hope Spots, Mission Blue, marine protected areas)", "text": "Summary clearly tied to Dr. Sylvia Earle's work or field of expertise (e.g., Hope Spots, Mission Blue, marine protected areas)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary is engaging and highlights why the event is worth attending", "text": "Summary is engaging and highlights why the event is worth attending", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "text": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Right Sidebar — Speaker Bio", "description": "The sidebar contains correct and verifiable speaker information.", "steps": [{"name": null, "description": "A shaded right-hand sidebar in **Seafoam Green** exists", "text": "A shaded right-hand sidebar in **Seafoam Green** exists", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Right sidebar contains text", "text": "Right sidebar contains text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speaker's current affiliation (e.g., National Geographic Explorer-in-Residence / Mission Blue) is explicitly named and correct", "text": "Speaker's current affiliation (e.g., National Geographic Explorer-in-Residence / Mission Blue) is explicitly named and correct", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Speaker's educational background (BS from Florida State, MS/PhD from Duke) is included and accurate", "text": "Speaker's educational background (BS from Florida State, MS/PhD from Duke) is included and accurate", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "3 distinct, verifiable professional achievements are listed (e.g., deepest sea walk, first female Chief Scientist of NOAA, Time Hero for the Planet)", "text": "3 distinct, verifiable professional achievements are listed (e.g., deepest sea walk, first female Chief Scientist of NOAA, Time Hero for the Planet)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sidebar includes connection between the speaker and Ocean Conservation", "text": "Sidebar includes connection between the speaker and Ocean Conservation", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Footer", "description": "The footer contains correct contact and reference information.", "steps": [{"name": null, "description": "Footer is positioned at the bottom right of the poster", "text": "Footer is positioned at the bottom right of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains contact info for the Scripps Institution of Oceanography at UC San Diego", "text": "Footer contains contact info for the Scripps Institution of Oceanography at UC San Diego", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains Dr. Sylvia Earle's webpage URL (missionblue.org)", "text": "Footer contains Dr. Sylvia Earle's webpage URL (missionblue.org)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Citations", "description": "All sources are properly cited and traceable.", "steps": [{"name": null, "description": "Citations are present in the speaker notes", "text": "Citations are present in the speaker notes", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sources cited correspond to specific claims in the topic summary", "text": "Sources cited correspond to specific claims in the topic summary", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sources cited correspond to specific claims in the speaker bio", "text": "Sources cited correspond to specific claims in the speaker bio", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "Overall Structure", "description": "All boxes have good structure.", "steps": [{"name": null, "description": "No text or images are overlapping each other", "text": "No text or images are overlapping each other", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are off of the slide", "text": "No text or images are off of the slide", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 37, "total_points": 185, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_3/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_3/checkpoints.md"}
{"task_id": "knows.slides_51_event_announcement_poster.4", "short_id": "51d", "family": "slides_51_event_announcement_poster", "instance": "instance_4", "instance_num": 4, "artifact": "Slides", "group": "Personal", "domain": "Party Invite / Event Poster", "prompt": "Create a one-page Google Slides formatted as a visually appealing event announcement poster. This is the Global Supply Chain Resilience Summit event, organized by the MIT Sloan School of Management. It will take place on 09/14/2026 at the Tang Center, Room E51-335 at MIT, Cambridge, MA. The main theme is how companies build resilient global supply chains, and the featured speaker is Yossi Sheffi.\nRetrieve reliable background information about how companies build resilient global supply chains that is closely related to Yossi Sheffi's work from a few stable, reputable sources (e.g., universities, professional associations, encyclopedias, major news outlets, and the speaker's website). Summarize the topic in 1-2 paragraphs that highlight why the event is worth attending and cite the websites you used. Also, introduce the speaker, including their education and current affiliation, and mention 3 of their main professional achievements.\nPoster Structure:\nHeader (top, centered): Global Supply Chain Resilience Summit, large, bold, in color Navy Blue.\nSubheader (below header, left-aligned): 09/14/2026, Tang Center Room E51-335, Cambridge, MA, and hosted by the MIT Sloan School of Management, styled in a smaller italic font.\nLogo (top, right): Retrieve the logo of the MIT Sloan School of Management and insert it to the top right of the poster. Make sure the logo size isn't too big or too small.\nCentral Section (main body, centered): Synthesized background on how companies build resilient global supply chains (engaging, persuasive, 1-2 short paragraphs).\nRight Sidebar (shaded box, color Charcoal Grey): Speaker bio (Yossi Sheffi), including his credentials, achievements, and relevance to global supply chain resilience.\nFooter (bottom, right-aligned): Contact info of the MIT Sloan School of Management, Yossi Sheffi's webpage.\nStyling Requirements: Ensure visual contrast between header, subheader, and body text. Use at least two colors: Navy Blue for header elements and Charcoal Grey for sidebars/highlights. Maintain consistent spacing and alignment so the document reads like a poster, not just plain text. Ensure that citations are placed in speaker notes and say exactly what they are citing in the poster.", "checkpoints": [{"n": 1, "title": "Header & Subheader", "description": "The header and subheader follow proper structure and contain correct information.", "steps": [{"name": null, "description": "Event name text box appears at the top of the page", "text": "Event name text box appears at the top of the page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text is centered on the slide", "text": "Event name text is centered on the slide", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has bold font", "text": "Event name text has bold font", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has font size >= 20", "text": "Event name text has font size >= 20", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text color is **Navy Blue**", "text": "Event name text color is **Navy Blue**", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name reads \"Global Supply Chain Resilience Summit\"", "text": "Event name reads \"Global Supply Chain Resilience Summit\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader exists below the main header", "text": "Subheader exists below the main header", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is left-aligned", "text": "Subheader is left-aligned", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is in italic font smaller than the main header", "text": "Subheader is in italic font smaller than the main header", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains date (09/14/2026)", "text": "Subheader contains date (09/14/2026)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains location (Tang Center Room E51-335, Cambridge, MA)", "text": "Subheader contains location (Tang Center Room E51-335, Cambridge, MA)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains host (MIT Sloan School of Management)", "text": "Subheader contains host (MIT Sloan School of Management)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Visual contrast is clear between header and subheader text", "text": "Visual contrast is clear between header and subheader text", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Logo", "description": "The logo is properly placed, scaled, and sourced.", "steps": [{"name": null, "description": "The logo is the **MIT Sloan School of Management** logo", "text": "The logo is the **MIT Sloan School of Management** logo", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is inserted in the top right of the slide", "text": "Logo is inserted in the top right of the slide", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo appropriately scaled (not disproportionately large or small)", "text": "Logo appropriately scaled (not disproportionately large or small)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo does not overlap or crowd header/subheader text", "text": "Logo does not overlap or crowd header/subheader text", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Central Body — Topic Background", "description": "The central body contains an accurate and well-sourced summary of the topic.", "steps": [{"name": null, "description": "Background summary is placed in the main, central body of the poster", "text": "Background summary is placed in the main, central body of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary is 1–2 paragraphs", "text": "Summary is 1–2 paragraphs", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary covers how companies build resilient global supply chains", "text": "Summary covers how companies build resilient global supply chains", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary clearly tied to Yossi Sheffi's work or field of expertise", "text": "Summary clearly tied to Yossi Sheffi's work or field of expertise", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary is engaging and highlights why the event is worth attending", "text": "Summary is engaging and highlights why the event is worth attending", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "text": "At least 2 reputable, stable sources (e.g., university, professional association, major news outlet, encyclopedia, or speaker's official website) are used", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Right Sidebar — Speaker Bio", "description": "The sidebar contains correct and verifiable speaker information.", "steps": [{"name": null, "description": "A shaded right-hand sidebar in **Charcoal Grey** exists", "text": "A shaded right-hand sidebar in **Charcoal Grey** exists", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Right sidebar contains text", "text": "Right sidebar contains text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speaker's current affiliation (MIT) is explicitly named and correct", "text": "Speaker's current affiliation (MIT) is explicitly named and correct", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Speaker's educational background (e.g., degrees from Technion / PhD from MIT) is included and accurate", "text": "Speaker's educational background (e.g., degrees from Technion / PhD from MIT) is included and accurate", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "3 distinct, verifiable professional achievements are listed", "text": "3 distinct, verifiable professional achievements are listed", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sidebar includes connection between the speaker and supply chain resilience", "text": "Sidebar includes connection between the speaker and supply chain resilience", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Footer", "description": "The footer contains correct contact and reference information.", "steps": [{"name": null, "description": "Footer is positioned at the bottom right of the poster", "text": "Footer is positioned at the bottom right of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains contact info for the MIT Sloan School of Management", "text": "Footer contains contact info for the MIT Sloan School of Management", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains Yossi Sheffi's webpage URL", "text": "Footer contains Yossi Sheffi's webpage URL", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Visual Styling and Layout", "description": "The document follows specific aesthetic requirements to look like an announcement poster.", "steps": [{"name": null, "description": "Color scheme uses Navy Blue for header text and Charcoal Grey for the sidebar/highlights", "text": "Color scheme uses Navy Blue for header text and Charcoal Grey for the sidebar/highlights", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are overlapping each other", "text": "No text or images are overlapping each other", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are off of the slide", "text": "No text or images are off of the slide", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 7, "title": "Citations and Speaker Notes", "description": "References are documented appropriately behind the scenes.", "steps": [{"name": null, "description": "Citations are present in the speaker notes", "text": "Citations are present in the speaker notes", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sources cited correspond to specific claims in the topic summary", "text": "Sources cited correspond to specific claims in the topic summary", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sources cited correspond to specific claims in the speaker bio", "text": "Sources cited correspond to specific claims in the speaker bio", "category": "content_check", "llm_vlm_primary": true}]}], "n_checkpoints": 7, "n_steps": 38, "total_points": 190, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_4/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_4/checkpoints.md"}
{"task_id": "knows.slides_51_event_announcement_poster.5", "short_id": "51e", "family": "slides_51_event_announcement_poster", "instance": "instance_5", "instance_num": 5, "artifact": "Slides", "group": "Personal", "domain": "Party Invite / Event Poster", "prompt": "Create a one-page Google Slides formatted as a visually appealing event announcement poster. This is the Breaking the Silence: Understanding the Youth Mental Health Crisis event, organized by the American Psychological Association. It will take place on 10/03/2026 at the Ronald Reagan Building and International Trade Center, Washington, D.C. The main theme is the growing mental health crisis among adolescents and young adults, and the featured speaker is Dr. Lisa Damour.\nRetrieve reliable background information about the growing mental health crisis among adolescents and young adults that is closely related to Dr. Lisa Damour's work from a few stable, reputable sources (e.g., universities, professional associations, encyclopedias, major news outlets, and the speaker's website). Summarize the topic in 1-2 paragraphs that highlight why the event is worth attending and cite the websites you used. Also, introduce the speaker, including their education and current affiliation, and mention 3 of their main professional achievements.\nPoster Structure:\nHeader (top, centered): Breaking the Silence: Understanding the Youth Mental Health Crisis, large, bold, in color Deep Teal.\nSubheader (below header, left-aligned): 10/03/2026, Ronald Reagan Building and International Trade Center, Washington, D.C., and hosted by the American Psychological Association, styled in a smaller italic font.\nLogo (top, right): Retrieve the logo of the American Psychological Association and insert it to the top right of the poster. Make sure the logo size isn't too big or too small.\nCentral Section (main body, centered): Synthesized background on the growing mental health crisis among adolescents and young adults (engaging, persuasive, 1-2 short paragraphs).\nRight Sidebar (shaded box, color Warm Amber): Speaker bio (Dr. Lisa Damour), including her credentials, achievements, and relevance to youth mental health.\nFooter (bottom, right-aligned): Contact info of the American Psychological Association, Dr. Lisa Damour's webpage.\nStyling Requirements: Ensure visual contrast between header, subheader, and body text. Use at least two colors: Deep Teal for header elements and Warm Amber for sidebars/highlights. Maintain consistent spacing and alignment so the document reads like a poster, not just plain text. Ensure that citations are placed in speaker notes and say exactly what they are citing in the poster.", "checkpoints": [{"n": 1, "title": "Header & Subheader", "description": "The header and subheader follow proper structure and contain correct information.", "steps": [{"name": null, "description": "Event name text box appears at the top of the page", "text": "Event name text box appears at the top of the page", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text is centered on the slide", "text": "Event name text is centered on the slide", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has bold font", "text": "Event name text has bold font", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text has font size >= 20", "text": "Event name text has font size >= 20", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name text color is **Deep Teal**", "text": "Event name text color is **Deep Teal**", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Event name reads \"Breaking the Silence: Understanding the Youth Mental Health Crisis\"", "text": "Event name reads \"Breaking the Silence: Understanding the Youth Mental Health Crisis\"", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader exists below the main header", "text": "Subheader exists below the main header", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is left-aligned", "text": "Subheader is left-aligned", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader is in italic font smaller than the main header", "text": "Subheader is in italic font smaller than the main header", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains date (10/03/2026)", "text": "Subheader contains date (10/03/2026)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains location (Ronald Reagan Building and International Trade Center, Washington, D.C.)", "text": "Subheader contains location (Ronald Reagan Building and International Trade Center, Washington, D.C.)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Subheader contains host (American Psychological Association)", "text": "Subheader contains host (American Psychological Association)", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Visual contrast is clear between header and subheader text", "text": "Visual contrast is clear between header and subheader text", "category": "formatting_check", "llm_vlm_primary": false}]}, {"n": 2, "title": "Logo", "description": "The official APA logo is properly placed, scaled, and sourced.", "steps": [{"name": null, "description": "The logo is the **American Psychological Association (APA)** logo", "text": "The logo is the **American Psychological Association (APA)** logo", "category": "visual_check", "llm_vlm_primary": true}, {"name": null, "description": "Logo is inserted in the top right of the slide", "text": "Logo is inserted in the top right of the slide", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo appropriately scaled (not pixelated, not obstructing text)", "text": "Logo appropriately scaled (not pixelated, not obstructing text)", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Logo does not overlap or crowd header/subheader text", "text": "Logo does not overlap or crowd header/subheader text", "category": "spatial_check", "llm_vlm_primary": false}]}, {"n": 3, "title": "Central Body — Topic Background", "description": "The central body provides a professional synthesis of the youth mental health crisis relevant to Dr. Damour's research.", "steps": [{"name": null, "description": "Background summary is placed in the main, central body of the poster", "text": "Background summary is placed in the main, central body of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary is 1–2 paragraphs", "text": "Summary is 1–2 paragraphs", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Summary covers the growing mental health crisis among adolescents and young adults", "text": "Summary covers the growing mental health crisis among adolescents and young adults", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary clearly tied to Dr. Lisa Damour's work or field of expertise", "text": "Summary clearly tied to Dr. Lisa Damour's work or field of expertise", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Summary is engaging and highlights why the event is worth attending", "text": "Summary is engaging and highlights why the event is worth attending", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "At least 2 reputable, stable sources (e.g., CDC, APA, university, or Dr. Damour's official website) are used", "text": "At least 2 reputable, stable sources (e.g., CDC, APA, university, or Dr. Damour's official website) are used", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 4, "title": "Right Sidebar — Speaker Bio", "description": "The sidebar contains correct and verifiable speaker information.", "steps": [{"name": null, "description": "A shaded right-hand sidebar in **Warm Amber** exists", "text": "A shaded right-hand sidebar in **Warm Amber** exists", "category": "formatting_check", "llm_vlm_primary": false}, {"name": null, "description": "Right sidebar contains text", "text": "Right sidebar contains text", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Speaker's current professional affiliation (e.g., Schubert Center for Child Studies at Case Western Reserve University) is mentioned and accurate", "text": "Speaker's current professional affiliation (e.g., Schubert Center for Child Studies at Case Western Reserve University) is mentioned and accurate", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Speaker's educational background (B.A. from Yale University, Ph.D. in Clinical Psychology from the University of Michigan) is included and accurate", "text": "Speaker's educational background (B.A. from Yale University, Ph.D. in Clinical Psychology from the University of Michigan) is included and accurate", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "3 distinct, verifiable professional achievements are listed (e.g., NYT bestselling author, *Inside Out 2* consultant, APA Thought Leader)", "text": "3 distinct, verifiable professional achievements are listed (e.g., NYT bestselling author, *Inside Out 2* consultant, APA Thought Leader)", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sidebar includes connection between the speaker and youth mental health", "text": "Sidebar includes connection between the speaker and youth mental health", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 5, "title": "Footer", "description": "The footer contains correct contact and reference information.", "steps": [{"name": null, "description": "Footer is positioned at the bottom right of the poster", "text": "Footer is positioned at the bottom right of the poster", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains contact info for the American Psychological Association", "text": "Footer contains contact info for the American Psychological Association", "category": "information_retrieval", "llm_vlm_primary": false}, {"name": null, "description": "Footer contains Dr. Lisa Damour's webpage URL (drlisadamour.com)", "text": "Footer contains Dr. Lisa Damour's webpage URL (drlisadamour.com)", "category": "information_retrieval", "llm_vlm_primary": false}]}, {"n": 6, "title": "Citations & Speaker Notes", "description": "References are documented appropriately behind the scenes.", "steps": [{"name": null, "description": "Citations are present in the speaker notes", "text": "Citations are present in the speaker notes", "category": "structural_check", "llm_vlm_primary": false}, {"name": null, "description": "Sources cited correspond to specific claims in the topic summary", "text": "Sources cited correspond to specific claims in the topic summary", "category": "content_check", "llm_vlm_primary": true}, {"name": null, "description": "Sources cited correspond to specific claims in the speaker bio", "text": "Sources cited correspond to specific claims in the speaker bio", "category": "content_check", "llm_vlm_primary": true}]}, {"n": 7, "title": "Overall Structure", "description": "All boxes have good structure.", "steps": [{"name": null, "description": "No text or images are overlapping each other", "text": "No text or images are overlapping each other", "category": "spatial_check", "llm_vlm_primary": false}, {"name": null, "description": "No text or images are off of the slide", "text": "No text or images are off of the slide", "category": "spatial_check", "llm_vlm_primary": false}]}], "n_checkpoints": 7, "n_steps": 37, "total_points": 185, "requires_browsing_history": true, "source_assets": [], "evaluator_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_5/evaluator.py", "checkpoints_url": "https://github.com/alexgill321/KNOWS-benchmark/blob/main/src/browsergym/knows/eval/tasks/slides_51_event_announcement_poster/instance_5/checkpoints.md"}