AlgoPuzzleVQA / metadata.json
ZTWHHH's picture
Add manifest (metadata.json)
9261d08 verified
Raw
History Blame Contribute Delete
1.84 kB
{
"name": "AlgoPuzzleVQA",
"release_date": "2024-03-21",
"subsets": {
"default": {
"language": [
"en"
],
"modalities": [
"single_image_start"
],
"task_type": "multiple_choice_qa",
"score_type": "rule",
"score_protocol": {
"reference": "official@github.com/declare-lab/LLM-PuzzleTest (AlgoPuzzleVQA shares the PuzzleVQA harness: main.py rule parse + ExactScorer exact match; no external LLM extractor/judge). Published rows confirm letter answers with options dict (e.g. answer 'B', options {A: Yes, B: No}).",
"note": "Official prompting is two-stage self-extraction by the evaluated model; scoring itself is rule-based exact match."
},
"prompt_template": "<image>{{ question }}\n{% for k, v in options.items() %}{{ k }}. {{ v }}\n{% endfor %}Answer with the option's letter from the given choices directly.",
"prompt_template_source": {
"origin": "official",
"reference": "https://huggingface.co/datasets/declare-lab/AlgoPuzzleVQA",
"notes": "Tier 1: AlgoPuzzleVQA single-image algorithmic-puzzle MCQ; options list, answer is option text mapped to letter."
},
"mapping_from_source": {
"media": {
"from": "images",
"type": "list",
"min_items": 1,
"max_items": 1
},
"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": {
"n_images": {
"from": "n_images"
},
"answer_text": {
"from": "answer_text"
}
},
"source": {
"format": "json",
"url": {
"data": "https://huggingface.co/datasets/declare-lab/AlgoPuzzleVQA"
}
}
}
}
}
}