MMT-Bench / metadata.json
ZTWHHH's picture
metadata: migrate score_type -> score_pipeline (atomic stage contract; see mm-eval scorer docs/en/SCORING.md)
4e034fc verified
Raw
History Blame Contribute Delete
2.72 kB
{
"name": "MMT-Bench",
"release_date": "2026-05-26",
"subsets": {
"main": {
"language": [
"en"
],
"modalities": [
"multi_image_start"
],
"task_type": "multiple_choice_qa",
"score_pipeline": [
"rule-match",
"llm-match"
],
"score_protocol": {
"reference": "VLMEvalKit@vlmeval/dataset/image_mcq.py:81-85,249-318 (MMT-Bench served by ImageMCQDataset.evaluate -> mcq_vanilla_eval; report_acc_MMT at :317-318) + vlmeval/dataset/utils/multiple_choice.py:359-406,475-508 — rule prefetch via can_infer (option letter / unique option-text), GPT extraction fallback, exact letter compare. Official MMT-Bench (OpenGVLab) evaluates via VLMEvalKit.",
"note": "Official headline aggregation is report_acc_MMT: accuracy grouped by category/l2_category (per-sample scorer yields overall accuracy only). mm-eval publishes the VAL split (3126 rows). lmms-eval mmt uses rule-only MMMU-style parsing (no GPT fallback) — VLMEvalKit protocol preferred as the authors' official harness."
},
"prompt_template": "{% for _ in range(n_images) %}<image>{% endfor %}{{ question }}\nOptions:\n{% for k, v in options.items() %}{{ k }}. {{ v }}\n{% endfor %}Please select the correct answer from the options above. ",
"mapping_from_source": {
"media": {
"from": "images",
"type": "list",
"min_items": 1,
"max_items": 23
},
"id": {
"from": "id"
},
"question": {
"from": "question"
},
"answer": {
"from": "answer",
"optional": true
},
"options": {
"from": "options",
"optional": true,
"note": "list source values are normalized to {A,B,...} dict"
},
"extra": {
"source_id": {
"from": "source_id"
},
"category": {
"from": "category"
},
"l2_category": {
"from": "l2_category"
},
"n_images": {
"from": "n_images"
}
},
"source": {
"format": "json",
"url": {
"val": "https://huggingface.co/datasets/Kaining/MMT-Bench"
}
}
},
"prompt_template_source": {
"origin": "official",
"reference": "https://github.com/EvolvingLMMs-Lab/lmms-eval/blob/main/lmms_eval/tasks/mmt/utils.py#L40-L55 (mmt_doc_to_text — multi-image MCQ with 'Question:/Options:/A./B./Please select the correct answer from the options above. ' format)",
"notes": "Tier 4: lmms-eval MMT-Bench canonical evaluation prompt."
}
}
}
}