metadata
dataset_info:
- config_name: art_and_design
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 3113808
num_examples: 11
- name: val
num_bytes: 25493074
num_examples: 88
- name: test
num_bytes: 311985416.171
num_examples: 1091
download_size: 343578142
dataset_size: 340592298.171
- config_name: business
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 554457
num_examples: 16
- name: val
num_bytes: 6152883
num_examples: 126
- name: test
num_bytes: 60103968.654
num_examples: 1538
download_size: 74809661
dataset_size: 66811308.654
- config_name: health_and_medicine
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 2837834
num_examples: 18
- name: val
num_bytes: 23957247
num_examples: 153
- name: test
num_bytes: 204211130.315
num_examples: 1865
download_size: 293138089
dataset_size: 231006211.315
- config_name: humanities_and_social_sciences
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 1406107
num_examples: 11
- name: val
num_bytes: 10657772
num_examples: 85
- name: test
num_bytes: 157998857.894
num_examples: 1038
download_size: 160096107
dataset_size: 170062736.894
- config_name: science
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 3750526
num_examples: 25
- name: val
num_bytes: 14762058
num_examples: 204
- name: test
num_bytes: 268839654.208
num_examples: 2494
download_size: 214377468
dataset_size: 287352238.208
- config_name: technology_and_engineering
features:
- name: id
dtype: string
- name: type
dtype: string
- name: source_type
dtype: string
- name: source
dtype: string
- name: question
dtype: string
- name: option1
dtype: string
- name: option2
dtype: string
- name: option3
dtype: string
- name: option4
dtype: string
- name: image_1
dtype: image
- name: image_2
dtype: image
- name: image_3
dtype: image
- name: image_4
dtype: image
- name: image_5
dtype: image
- name: answer
dtype: string
- name: analysis
dtype: string
- name: distribution
dtype: string
- name: difficulty_level
dtype: string
- name: subcategory
dtype: string
- name: category
dtype: string
- name: subfield
dtype: string
- name: img_type
dtype: string
- name: image_1_filename
dtype: string
- name: image_2_filename
dtype: string
- name: image_3_filename
dtype: string
- name: image_4_filename
dtype: string
- name: image_5_filename
dtype: string
splits:
- name: dev
num_bytes: 1518231
num_examples: 31
- name: val
num_bytes: 14794870
num_examples: 244
- name: test
num_bytes: 134814024.93
num_examples: 2974
download_size: 213411677
dataset_size: 151127125.93
configs:
- config_name: art_and_design
data_files:
- split: dev
path: art_and_design/dev-*
- split: val
path: art_and_design/val-*
- split: test
path: art_and_design/test-*
- config_name: business
data_files:
- split: dev
path: business/dev-*
- split: val
path: business/val-*
- split: test
path: business/test-*
- config_name: health_and_medicine
data_files:
- split: dev
path: health_and_medicine/dev-*
- split: val
path: health_and_medicine/val-*
- split: test
path: health_and_medicine/test-*
- config_name: humanities_and_social_sciences
data_files:
- split: dev
path: humanities_and_social_sciences/dev-*
- split: val
path: humanities_and_social_sciences/val-*
- split: test
path: humanities_and_social_sciences/test-*
- config_name: science
data_files:
- split: dev
path: science/dev-*
- split: val
path: science/val-*
- split: test
path: science/test-*
- config_name: technology_and_engineering
data_files:
- split: dev
path: technology_and_engineering/dev-*
- split: val
path: technology_and_engineering/val-*
- split: test
path: technology_and_engineering/test-*
CMMMU
🌐 Homepage | 🤗 Paper | 📖 arXiv | 🤗 Dataset | GitHub
Introduction
CMMMU includes 12k manually collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering, like its companion, MMMU. These questions span 30 subjects and comprise 39 highly heterogeneous image types, such as charts, diagrams, maps, tables, music sheets, and chemical structures.
🏆 Mini-Leaderboard
| Model | Val (900) | Test (11K) |
|---|---|---|
| GPT-4V(ision) (Playground) | 42.5 | 43.7 |
| Qwen-VL-PLUS* | 39.5 | 36.8 |
| Yi-VL-34B | 36.2 | 36.5 |
| Yi-VL-6B | 35.8 | 35.0 |
| Qwen-VL-7B-Chat | 30.7 | 31.3 |
| InternVL-Chat-ViT-6B-Vicuna-7B | 26.4 | 26.7 |
| InternVL-Chat-ViT-6B-Vicuna-13B | 27.4 | 26.1 |
| CogAgent-Chat | 24.6 | 23.6 |
| Emu2-Chat | 23.8 | 24.5 |
| Chinese-LLaVA | 25.5 | 23.4 |
| VisCPM | 25.2 | 22.7 |
| mPLUG-OWL2 | 20.8 | 22.2 |
| Frequent Choice | 24.1 | 26.0 |
| Random Choice | 21.6 | 21.6 |
*: results provided by the authors.
Disclaimers
The guidelines for the annotators emphasized strict compliance with copyright and licensing rules from the initial data source, specifically avoiding materials from websites that forbid copying and redistribution. Should you encounter any data samples potentially breaching the copyright or licensing regulations of any site, we encourage you to contact us. Upon verification, such samples will be promptly removed.
Contact
- Ge Zhang: zhangge@01.ai
- Wenhao Huang: huangwenhao@01.ai
- Xinrun Du: duxinrun@01.ai
- Bei Chen: chenbei@01.ai
- Wenhu Chen: wenhuchen@uwaterloo.ca
- Jie Fu: jiefu@ust.hk
Citation
BibTeX:
@article{zhang2024cmmmu,
title={CMMMU: A Chinese Massive Multi-discipline Multimodal Understanding Benchmark},
author={Ge, Zhang and Xinrun, Du and Bei, Chen and Yiming, Liang and Tongxu, Luo and Tianyu, Zheng and Kang, Zhu and Yuyang, Cheng and Chunpu, Xu and Shuyue, Guo and Haoran, Zhang and Xingwei, Qu and Junjie, Wang and Ruibin, Yuan and Yizhi, Li and Zekun, Wang and Yudong, Liu and Yu-Hsuan, Tsai and Fengji, Zhang and Chenghua, Lin and Wenhao, Huang and Wenhu, Chen and Jie, Fu},
journal={arXiv preprint arXiv:2401.20847},
year={2024},
}