VSI-Bench / metadata.json
ZTWHHH's picture
metadata: migrate score_type -> score_pipeline (atomic stage contract; see mm-eval scorer docs/en/SCORING.md)
dbf2eb5 verified
Raw
History Blame Contribute Delete
3.44 kB
{
"name": "VSI-Bench",
"release_date": "2024-12-18",
"subsets": {
"mc": {
"language": [
"en"
],
"modalities": [
"single_video_start"
],
"task_type": "multiple_choice_qa",
"score_pipeline": [
"exact-match",
"rule-match"
],
"score_protocol": {
"reference": "lmms-eval@lmms_eval/tasks/vsibench/utils.py:87-89,115-121 — MCA question types graded by fuzzy_matching (first token, strip trailing period) + case-insensitive exact match against the ground-truth letter; no LLM.",
"note": "The org ships ONLY the multiple-choice question types (mc config, 2490 rows; confirmed in rows file). The official benchmark also contains NA (numerical) types graded with Mean Relative Accuracy MRA:.5:.95:.05 (utils.py:95-100,29-31), and the official headline 'overall' averages 8 per-question-type scores after averaging object_rel_direction easy/medium/hard (utils.py:158-170) — neither is reproducible from this copy or from a per-sample mean. Rows carry extra 'question_type' with VSI type names (object_rel_direction_hard, ...) that are not scorer vocabulary."
},
"prompt_template": "<video>These are frames of a video.\n{{ question }}\nOptions:\n{% for o in options %}{{ o }}\n{% endfor %}Answer with the option's letter from the given choices directly.",
"prompt_template_source": {
"origin": "official",
"reference": "https://github.com/EvolvingLMMs-Lab/lmms-eval/blob/main/lmms_eval/tasks/vsibench/utils.py (vsibench_doc_to_text — canonical MCA prompt) + _default_template_yaml",
"notes": "Tier 4: lmms-eval VSI-Bench canonical MCA prompt for the MC question types. SAT paper (arXiv 2412.07755, Appendix A.3, p.20) restricts VSI-Bench evaluation to the MC split — numerical splits (object_counting, object_size_estimation, room_size_estimation, object_abs_distance) are excluded. Each MC row carries the source `options` list verbatim from nyu-visionx/VSI-Bench under `choices`."
},
"video_storage": {
"format": "files",
"media_root": "videos",
"notes": "512 source videos (243 referenced by the MC split): arkitscenes (121), scannet (79), scannetpp (43). Videos are mirrored from lccshunli/VSIBench at 5.7 GB total; nyu-visionx/VSI-Bench provides only QA metadata. Each scene_name resolves to videos/<dataset_source>/<scene_name>.mp4. Frame sampling is the inference backend's responsibility."
},
"mapping_from_source": {
"media": {
"from": "video_path",
"type": "list",
"min_items": 1,
"max_items": 1
},
"id": {
"from": "id"
},
"question": {
"from": "question"
},
"answer": {
"from": "ground_truth"
},
"choices": {
"from": "options"
},
"extra": {
"scene_name": {
"from": "scene_name"
},
"dataset_source": {
"from": "dataset"
},
"question_type": {
"from": "question_type"
}
},
"source": {
"format": "huggingface",
"url": {
"test": "https://huggingface.co/datasets/nyu-visionx/VSI-Bench (QA metadata, 'full' config); videos mirrored from https://huggingface.co/datasets/lccshunli/VSIBench"
}
}
}
}
}
}