TP Bench – Theoretical Physics Benchmark for AI

TPBench is a curated dataset and evaluation suite designed to measure the reasoning capabilities of AI models in theoretical physics. Our test problems span multiple difficulty levels—from undergraduate to frontier research—and cover topics such as cosmology, high-energy theory, general relativity, and more. By providing a unified framework for problem-solving and auto-verifiable answers, TPBench aims to drive progress in AI-based research assistance for theoretical physics.

Dataset Sources

• Paper: https://arxiv.org/abs/2502.15815
• Website: https://tpbench.org/

Citation

If you do find our dataset helpful, please cite our paper.

BibTeX:

@misc{chung2025theoreticalphysicsbenchmarktpbench,  
  title={Theoretical Physics Benchmark (TPBench) -- a Dataset and Study of AI Reasoning Capabilities in Theoretical Physics},   
  author={Daniel J. H. Chung and Zhiqi Gao and Yurii Kvasiuk and Tianyi Li and Moritz Münchmeyer and Maja Rudolph and Frederic Sala and Sai Chaitanya Tadepalli},  
  year={2025},  
  eprint={2502.15815},  
  archivePrefix={arXiv},  
  primaryClass={cs.LG},  
  url={https://arxiv.org/abs/2502.15815},   
}

Dataset Card Contact

Please email us at research@tpbench.org if you have questions or concerns.