MM-K12

[📂 GitHub] [📜 Paper]

MM-K12 is a curated, high-quality dataset containing 10,000 multimodal math problems sourced from K-12 educational content. Each problem includes both textual and visual components, covering a wide range of mathematical topics (e.g., arithmetic, geometry, algebra). All problems have unique, verifiable answers, making the dataset ideal for supervised training, evaluation, and reward modeling in multimodal mathematical reasoning tasks.

The dataset serves as seed data for the automatic generation of step-level supervision within the MM-PRM framework. An additional 500 examples are reserved as a test set.

Data fields

Citation

If you find this project useful in your research, please consider citing:

@article{du2025mmprm,
      title={MM-PRM: Enhancing Multimodal Mathematical Reasoning with Scalable Step-Level Supervision},
      author={Lingxiao Du and Fanqing Meng and Zongkai Liu and Zhixiang Zhou and Ping Luo and Qiaosheng Zhang and Wenqi Shao},
      year={2025},
      journal={arXiv preprint arXiv:2505.13427},
}