Lean4Physics: Comprehensive Reasoning Framework for College-level Physics in Lean4
Yuxin Li, Minghao Liu, Ruida Wang, Wenzhao Ji, Zhitao He, Rui Pan, Junming Huang, Tong Zhang, Yi R. Fung
摘要
We present Lean4PHYS, a comprehensive reasoning framework for college-level physics problems in Lean4. To establish a solid foundation for formal reasoning in physics, Lean4PHYS launches PhysLib, a repository containing fundamental unit systems and essential theorems to formulate physics proofs in Lean4. It will be community-driven and long-term maintained. Lean4PHYS also includes LeanPhysBench, a college-level benchmark for evaluating LLMs' Lean4 formal physics reasoning capability. It contains 200 hand-crafted and peer-reviewed Lean4 theorem statements formalized from university textbooks and physics competition problems. Based on the PhysLib and LeanPhysBench we composed in Lean4PHYS, we perform exhaustive experiments of baseline results using major expert Math provers and state-of-the-art closed-source models, and provide an analysis of their performance. In the experiment, we identify that most expert provers do not outperform general models as they did in the math domain. This suggests potential overfitting to the math domain rather than learning formal reasoning for formal provers. We also conduct a comprehensive experiment showing that, with PhysLib in the context, LLMs' performance on LeanPhysBench increases by 11.90% on average, proving the effectiveness of our repository in assisting LLMs in solving the Lean4 physics problem. To the best of our knowledge, we are the first study to provide a physics benchmark in Lean4.
问问这篇 Paper
智能体会读完全文。
Lune 把这篇 Paper 索引到了每一个公式,引用它的顶会 Paper 也一样。你提问,回答直接引用原文。
引用它的顶会 Paper2
- Dancing in Chains: Strategic Persuasion in Academic Rebuttal via Theory of MindZhitao He, Zongwei Lyu, Yi R. FungICLR 2026 · 被引用 5 次
- GAR: Generative Adversarial Reinforcement Learning for Formal Theorem ProvingRuida Wang, Jiarui Yao, Rui Pan, Shizhe Diao 等ICLR 2026 · 被引用 5 次
它引用的顶会 Paper13
- miniF2F: a cross-system benchmark for formal Olympiad-level mathematicsKunhao Zheng, Jesse Michael Han, Stanislas PoluICLR 2022 · 被引用 342 次
- Goedel-Prover-V2: Scaling Formal Theorem Proving with Scaffolded Data Synthesis and Self-CorrectionYong Lin, Shange Tang, Bohan Lyu, Ziran Yang 等ICLR 2026 · 被引用 160 次
- BFS-Prover: Scalable Best-First Tree Search for LLM-based Automatic Theorem ProvingRan Xin, Chenguang Xi, Jie Yang, Feng Chen 等ACL 2025 · 被引用 66 次
- PhysReason: A Comprehensive Benchmark towards Physics-Based ReasoningXinyu Zhang, Yuxuan Dong, Yanrui Wu, Jiaxing Huang 等ACL 2025 · 被引用 51 次
- Formal Mathematics Statement Curriculum LearningStanislas Polu, Jesse Michael Han, Kunhao Zheng, Mantas Baksys 等ICLR 2023 · 被引用 24 次
相关 Paper
- UGPhysics: A Comprehensive Benchmark for Undergraduate Physics Reasoning with Large Language ModelsXin Xu, Qiyun Xu, Tong Xiao, Tianhao Chen 等ICML 2025
- SAIR-Comb : A Structure-Aware Iterative Refinement Framework for Combinatorics AutoformalizationWeijie Jiang, Gaolei He, Beibei Xiong, Jianlin Wang 等ACL 2026
- Hilbert: Recursively Building Formal Proofs with Informal ReasoningSumanth Varambally, Thomas Voice, Yanchao Sun, Zhifeng Chen 等ICLR 2026 · 被引用 62 次
- FANS: Formal Answer Selection for LLM Natural Language Math Reasoning Using Lean4Jiarui Yao, Ruida Wang, Tong ZhangEMNLP 2025 · 被引用 1 次
- CriticLean: Critic-Guided Reinforcement Learning for Mathematical FormalizationZhongyuan Peng, Yifan Yao, Kaijing Ma, Shuyue Guo 等ACL 2026 · 被引用 15 次
