Datasets:

GenuineWWD
/

SCS_data

Self‑Consistency Sampling (SCS) improves outcome‑reward reinforcement learning for multimodal large language models (MLLMs). In multiple‑choice reasoning tasks, models often get the correct answer through faulty reasoning and receive unmerited rewards. SCS mitigates this by introducing visual perturbations and repeated resampling of reasoning trajectories, rewarding only consistent reasoning paths. Integrated into methods like RLOO, GRPO, and REINFORCE++, SCS boosts accuracy by up to 7.7% on six multimodal benchmarks with minimal extra cost, and generalizes across models including Qwen2.5‑VL and InternVL3.

Training

Please refer to code repo for more details.

Evaluation

Please refer to code repo for more details.

Contact

Jiahao Wang: [email protected]
Weiye Xu: [email protected]

Citation

BibTeX:

@article{wang2025enhancing,
  title={Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling},
  author={Wang, Jiahao and Xu, Weiye and Yang, Aijun and Zhou, Wengang and Lu, Lewei and Li, Houqiang and Wang, Xiaohua and Zhu, Jinguo},
  journal={arXiv preprint arXiv:2511.10648},
  year={2025}
}

Downloads last month: 44

Size of downloaded dataset files:

25.5 MB

Size of the auto-converted Parquet files:

5.01 MB

Number of rows:

32,360

id stringlengths 1 25 ⌀	question stringlengths 7 682	answer stringclasses 5 values	message stringlengths 388 1.3k
null	What is the name of the colony shown?	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/2027/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhat is the name of the colony shown?A. Florida\nB. Washington, D.C.\nC. South Carolina\nD. Maryland"}]}]
null	Is chalcopyrite a mineral?	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/1697/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nIs chalcopyrite a mineral?A. no\nB. yes"}]}]
physical-commonsense-1435	What is the purpose of the tower?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-physical-commonsense-407.png"}, {"type": "text", "text": "<image>\nWhat is the purpose of the tower?\nOptions:\nA. To provide shade\nB. Only to indicate the time\nC. To indicate the time and wind direction\nD. To store water\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
economics-662	I need the average hourly wages information for Country 2 and Country 5 in 2022. Can you provide it?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/social-science-economics-1713.png"}, {"type": "text", "text": "<image>\nI need the average hourly wages information for Country 2 and Country 5 in 2022. Can you provide it?\nOptions:\nA. 61.53 ($/h), 122.47 ($/h)\nB. 49.60 ($/h), 122.47 ($/h)\nC. 122.47 ($/h), 61.53 ($/h)\nD. 122.47 ($/h), 49.60 ($/h)\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
2006	Find the area of the figure. Round to the nearest tenth if necessary.	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/2006/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind the area of the figure. Round to the nearest tenth if necessary.\nOptions:\nA. 36\nB. 120\nC. 156\nD. 192"}]}]
physical-commonsense-10	Among the objects listed in the image's options, which ones will ultimately sink when placed in water?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/natural-science-physics-5011.png"}, {"type": "text", "text": "<image>\nAmong the objects listed in the image's options, which ones will ultimately sink when placed in water?\nOptions:\nA. 0\nB. 1\nC. 2\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
biology-19	Based on this information, what is this pea plant's genotype for the stem height gene?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/natural-science-biology-20.png"}, {"type": "text", "text": "<image>\nBased on this information, what is this pea plant's genotype for the stem height gene?\nOptions:\nA. HH\nB. Hh\nC. hh\nD. aa\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
physics-1046	Think about the magnetic force between the magnets in each pair. Which of the following statements is true?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/natural-science-physics-1046.png"}, {"type": "text", "text": "<image>\nThink about the magnetic force between the magnets in each pair. Which of the following statements is true?\nOptions:\nA. The magnetic force is stronger in Pair 2.\nB. The magnetic force is stronger in Pair 1.\nC. The strength of the magnetic force is the same in both pairs.\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
writing-strategies-83	Which of the following contains a vague pronoun reference?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/language-science-writing-strategies-111.png"}, {"type": "text", "text": "<image>\nWhich of the following contains a vague pronoun reference?\nOptions:\nA. Option A in the image\nB. Option B in the image\nC. The given choices are misleading.\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
null	What is the capital of California?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/8026/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhat is the capital of California?A. Sacramento\nB. San Francisco\nC. Los Angeles\nD. Santa Fe"}]}]
physics-1359	Which property do these three objects have in common?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-physical-commonsense-5359.png"}, {"type": "text", "text": "<image>\nWhich property do these three objects have in common?\nOptions:\nA. sour\nB. blue\nC. salty\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
geography-3496	Can you tell me the height of the northernmost point?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/social-science-geography-3496.png"}, {"type": "text", "text": "<image>\nCan you tell me the height of the northernmost point?\nOptions:\nA. 4.9 km\nB. 3.8 km\nC. -1.3 km\nD. 1.1 km\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
temporal-commonsense-113	What season is likely when the picture was taken?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-temporal-commonsense-464.png"}, {"type": "text", "text": "<image>\nWhat season is likely when the picture was taken?\nOptions:\nA. Spring\nB. Summer\nC. Fall\nD. Winter\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
null	What is the capital of Maryland?	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/6255/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhat is the capital of Maryland?A. Augusta\nB. Annapolis\nC. Baltimore\nD. Biloxi"}]}]
null	Which letter marks the Louisiana territory?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/2630/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich letter marks the Louisiana territory?A. A\nB. D\nC. C\nD. B"}]}]
null	Which of these states is farthest west?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/17799/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich of these states is farthest west?A. Kentucky\nB. Michigan\nC. Montana\nD. Florida"}]}]
mathematics-738	The following picture is a test of a Geometry problem. Which of the following choices given in the text is correct?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/mathematics-geometry-5389.png"}, {"type": "text", "text": "<image>\nThe following picture is a test of a Geometry problem. Which of the following choices given in the text is correct?\nOptions:\nA. None of the choices given in the text are correct.\nB. option C in the image is the correct answer.\nC. the answer is option D in the image\nD. the solution is option A in the image.\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
null	Which of these states is farthest east?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/3734/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich of these states is farthest east?A. South Dakota\nB. Washington\nC. California\nD. Utah"}]}]
null	Which solution has a higher concentration of green particles?	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/10009/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich solution has a higher concentration of green particles?A. neither; their concentrations are the same\nB. Solution B\nC. Solution A"}]}]
physical-commonsense-100	In the provided image, how many items from the given choices will submerge underwater?	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/natural-science-physics-5133.png"}, {"type": "text", "text": "<image>\nIn the provided image, how many items from the given choices will submerge underwater?\nOptions:\nA. 3\nB. 0\nC. 1\nD. 2\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
2037	Find y.	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/2037/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind y.\nOptions:\nA. 1.5\nB. 2\nC. 2.5\nD. 3"}]}]
economics-336	Can you tell me the count of countries with fiscal surpluses in 2016?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/social-science-economics-1359.png"}, {"type": "text", "text": "<image>\nCan you tell me the count of countries with fiscal surpluses in 2016?\nOptions:\nA. 7\nB. 5\nC. 2\nD. 6\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
144	Find x.	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/144/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind x.\nOptions:\nA. 7\nB. 12\nC. 22\nD. 31"}]}]
mathematics-515	If you had to select one option as the correct answer for the Number Theory problem shown in the picture, which one would you choose and why?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/mathematics-theory-4946.png"}, {"type": "text", "text": "<image>\nIf you had to select one option as the correct answer for the Number Theory problem shown in the picture, which one would you choose and why?\nOptions:\nA. option C in the image should be selected.\nB. All given choices given in the text are wrong.\nC. the solution is option B in the image.\nD. the answer is option A in the image\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
null	Which country is highlighted?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/12127/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich country is highlighted?A. Barbados\nB. the Dominican Republic\nC. Saint Lucia\nD. Saint Vincent and the Grenadines"}]}]
0	\triangle R S T \cong \triangle X Y Z. Find y.	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/0/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\n\\triangle R S T \\cong \\triangle X Y Z. Find y.\nOptions:\nA. 5\nB. 14\nC. 15\nD. 35"}]}]
social-commonsense-1412	What kind of area is shown in the picture?	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-social-commonsense-1016.png"}, {"type": "text", "text": "<image>\nWhat kind of area is shown in the picture?\nOptions:\nA. A park meant for dogs\nB. A nature reserve\nC. A lake meant for swimming\nD. A park meant for ducks\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
social-commonsense-1334	Why is the man carrying a suitcase on the beach?	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-social-commonsense-921.png"}, {"type": "text", "text": "<image>\nWhy is the man carrying a suitcase on the beach?\nOptions:\nA. He is going on a vacation\nB. He is a lifeguard on duty\nC. He is delivering something to the beach\nD. He is working nearby and has to ask a question\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
2048	Find x so that the quadrilateral is a parallelogram.	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/2048/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind x so that the quadrilateral is a parallelogram.\nOptions:\nA. 28\nB. 30\nC. 90\nD. 112"}]}]
mathematics-951	After reviewing the following questions, can you determine if option A in the image is the correct choice?	B	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/mathematics-geometry-5627.png"}, {"type": "text", "text": "<image>\nAfter reviewing the following questions, can you determine if option A in the image is the correct choice?\nOptions:\nA. Not sure\nB. True\nC. False\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
657	Find x.	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/657/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind x.\nOptions:\nA. 1.5\nB. 2.0\nC. 8.0\nD. 8.5"}]}]
grammar-74	Which is a compound sentence?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/language-science-grammar-144.png"}, {"type": "text", "text": "<image>\nWhich is a compound sentence?\nOptions:\nA. Option A in the image\nB. Option B in the image\nC. It seems like there's an error in all the provided options.\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
1577	Find x. Round to the nearest hundredth.	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/1577/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind x. Round to the nearest hundredth.\nOptions:\nA. 4.47\nB. 5.34\nC. 6.54\nD. 7.21"}]}]
null	Which type of force from the hammer moves the nail into the wood?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/19810/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nWhich type of force from the hammer moves the nail into the wood?A. push\nB. pull"}]}]
null	Think about the magnetic force between the magnets in each pair. Which of the following statements is true?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/4035/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nThink about the magnetic force between the magnets in each pair. Which of the following statements is true?A. The magnitude of the magnetic force is greater in Pair 2.\nB. The magnitude of the magnetic force is the same in both pairs.\nC. The magnitude of the magnetic force is greater in Pair 1."}]}]
1065	Find the area of the shaded sector. Round to the nearest tenth, if necessary.	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "geometry3k/train/1065/img_diagram.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nFind the area of the shaded sector. Round to the nearest tenth, if necessary.\nOptions:\nA. 10.8\nB. 59.6\nC. 60.2\nD. 333.9"}]}]
mathematics-1498	Look at the following schedule. When does the badminton event begin?	D	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-temporal-commonsense-2630.png"}, {"type": "text", "text": "<image>\nLook at the following schedule. When does the badminton event begin?\nOptions:\nA. 3:55 PM\nB. 2:55 PM\nC. 3:00 PM\nD. 1:20 PM\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
null	Select the amphibian below.	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "scienceqa/train/17262/image.png"}, {"type": "text", "text": "<image>\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>.\nSelect the amphibian below.A. olive toad\nB. bison"}]}]
geography-3795	What are the elevations of the points Point B, Point C and Point D individually?	A	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/social-science-geography-3795.png"}, {"type": "text", "text": "<image>\nWhat are the elevations of the points Point B, Point C and Point D individually?\nOptions:\nA. 6.8, 9.7, -5.7\nB. 6.8, 6.5, -5.7\nC. 6.8, 9.7, 3.4\nD. 9.7, 6.8, -5.7\nE. 6.8, -2.4, -5.7\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]
physics-516	Which property do these three objects have in common?	C	[{"role": "system", "content": "你是书生·万象，英文名是InternVL，是由上海人工智能实验室、清华大学及多家合作单位联合开发的多模态大语言模型。"}, {"role": "user", "content": [{"type": "image", "image": "m3cot_own/images/commonsense-physical-commonsense-4516.png"}, {"type": "text", "text": "<image>\nWhich property do these three objects have in common?\nOptions:\nA. transparent\nB. rough\nC. flexible\nSolve the problem through step-by-step reasoning and answer directly with the option letter. Think about the reasoning process first and answer the question following this format: <think> THINKING </think><answer> ANSWER </answer>."}]}]