NExT-QA / metadata.json
ZTWHHH's picture
Tighten prompt_template_source.notes: explicit 5-tier classification per 2026-06-01 priority rules
4214884 verified
Raw
History Blame Contribute Delete
1.96 kB
{
"name": "NExT-QA",
"release_date": "2021-05-01",
"subsets": {
"mc": {
"language": [
"en"
],
"modalities": [
"single_video_start"
],
"task_type": "multiple_choice_vqa",
"prompt_template": "<video>{{ question }}\n{% for k, v in options.items() %}{{ k }}. {{ v }}\n{% endfor %}Answer with the option's letter from the given choices directly.",
"video_storage": {
"format": "files",
"media_root": "videos",
"notes": "1,000 videos from VidOR dataset (out of 1,570 total in source). MC test split: 8,564 QA pairs across causal, temporal, and descriptive types."
},
"mapping_from_source": {
"source": {
"format": "huggingface",
"url": {
"train": "https://huggingface.co/datasets/lmms-lab/NExTQA"
}
},
"id": {
"from": "video+qid",
"note": "Composite key MC_{video_id}_{qid}"
},
"question": {
"from": "question"
},
"options": {
"from": "a0-a4",
"note": "5 choices normalized to {A,...,E} dict"
},
"answer": {
"from": "answer",
"note": "Integer index converted to letter"
},
"media": {
"from": "video",
"type": "list",
"min_items": 1,
"max_items": 1
},
"extra": {
"video_id": {
"from": "video"
},
"question_type": {
"from": "type"
}
}
},
"prompt_template_source": {
"origin": "official",
"reference": "https://github.com/EvolvingLMMs-Lab/lmms-eval/blob/main/lmms_eval/tasks/nextqa/utils.py (nextqa_mc_doc_to_text — A./B./C./D./E. options + 'Answer with the option's letter from the given choices directly.')",
"notes": "Tier 4: lmms-eval NExT-QA MC canonical evaluation prompt."
}
}
}
}