README.md

metadata

language:
  - en
license: mit
task_categories:
  - text-generation
tags:
  - survey
  - mathematical-reasoning
  - geometry
  - deep-learning

A Survey of Deep Learning for Geometry Problem Solving

This repository is the reading list accompanying the paper A Survey of Deep Learning for Geometry Problem Solving.

It aims to provide a comprehensive and practical reference for deep learning applications in geometry problem solving. The repository is continuously updated and organized to allow searching for corresponding papers by year.

Paper Abstract: Recent years have seen rapid advancements in deep learning, especially multimodal large language models, sparking a research boom in geometry problem solving—a critical area for mathematical reasoning, AI assessment, and multimodal capabilities. This paper surveys deep learning applications in this field, summarizing relevant tasks, reviewing deep learning methods, analyzing evaluation metrics, and discussing challenges and future directions. Our goal is to offer a comprehensive, practical reference to foster further progress. We maintain a continuously updated paper list on GitHub.

GitHub Repository: https://github.com/majianz/gps-survey

The reading list includes papers categorized by:

• Surveys
• Tasks and Datasets (Fundamental, Core, Composite, and Other Geometry Tasks)
• Architectures
• Methods
• Related Surveys

Citation

If you find this survey helpful in your research, please consider citing the paper:

@misc{ma2025survey,
      title={A Survey of Deep Learning for Geometry Problem Solving}, 
      author={Jianzhe Ma and Tianyi Li and Haoran Wei and Yingqian Min and Lei Fang and Ji-Rong Wen},
      year={2025},
      eprint={2507.11936},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}