WHOOPS / metadata.json
ZTWHHH's picture
metadata: migrate score_type -> score_pipeline (atomic stage contract; see mm-eval scorer docs/en/SCORING.md)
d2bfb84 verified
Raw
History Blame Contribute Delete
2.54 kB
{
"name": "WHOOPS!",
"release_date": "2023-03-13",
"subsets": {
"main": {
"language": [
"en"
],
"modalities": [
"single_image_start"
],
"task_type": "short_answer_qa",
"score_pipeline": [
"exact-match",
"rule-match",
"llm-judge"
],
"score_protocol": {
"reference": "WHOOPS! paper (Bitton-Guetta et al., ICCV 2023; arxiv 2303.07274) — the VQA sub-task is graded with BEM (BERT-based answer-equivalence classifier judging prediction vs reference correct/incorrect), reported alongside exact match. mm-eval ships the VQA sub-task (WHOOPS.json explodes `qa` into question/answer).",
"note": "Official headline metric is BEM (learned BERT answer-equivalence), not a rule or an LLM judge. Modeled as rule_llm_judge (exact match short-circuit + LLM judge for semantic equivalence) as the closest reproducible analog; results will differ numerically from BEM. WHOOPS also has captioning/explanation/matching sub-tasks not shipped here."
},
"prompt_template": "<image>{{ question }}\nAnswer the question using a single word or phrase.",
"mapping_from_source": {
"media": {
"from": "image",
"type": "list",
"min_items": 1,
"max_items": 1
},
"id": {
"from": "id",
"explode_from": "qa",
"id_template": "{id}_q{idx}"
},
"question": {
"from": "question"
},
"answer": {
"from": "answer",
"optional": true
},
"extra": {
"commonsense_category": {
"from": "commonsense_category"
},
"designer_explanation": {
"from": "designer_explanation"
}
},
"source": {
"format": "huggingface",
"url": {
"test": "https://huggingface.co/datasets/nlphuji/whoops"
}
}
},
"prompt_template_source": {
"origin": "fallback",
"reference": "T1",
"notes": "Tier 5: no higher tier publishes a prompt. T1: whoops-benchmark.github.io and the ICCV 2023 paper publish no reusable model-input prompt (paper VQA numbers use BLIP2 whose 'Question: {} Short answer:' is BLIP2's own convention, not a WHOOPS-authored prompt); T2: no prompt column in nlphuji/whoops; T3: no VLMEvalKit whoops; T4: no lmms-eval tasks/whoops (clone @047ec52). Template is the canonical single-image short-answer fallback."
}
}
}
}