Datasets:
metadata
license: bsd-3-clause
language:
- zh
pipeline_tag: text-generation
CMM-Math
Introduction
We release the Chinese Multimodal Mathemathical Dataset (CMM-Math), which contains 22k+ training samples and 5k+ evaluation samples. More details can be found in our following paper:
@article{liu2024cmm,
title={CMM-Math: A Chinese Multimodal Math Dataset To Evaluate and Enhance the Mathematics Reasoning of Large Multimodal Models},
author={Liu, Wentao and Pan, Qianjun and Zhang, Yi and Liu, Zhuo and Wu, Ji and Zhou, Jie and Zhou, Aimin and Chen, Qin and Jiang, Bo and He, Liang},
journal={arXiv preprint arXiv:2409.02834},
year={2024}
}
Demo
Some examples in our dataset
The results over different subjects
The results over different levels
