Saved in:
Bibliographic Details
Main Authors: Xuejun, Yu, Zhong, Jianyuan, Feng, Zijin, Zhai, Pengyi, Yousefzadeh, Roozbeh, Ng, Wei Chong, Liu, Haoxiong, Shou, Ziyi, Xiong, Jing, Zhou, Yudong, Ong, Claudia Beth, Sugiarto, Austen Jeremy, Zhang, Yaoxi, Tai, Wai Ming, Cao, Huan, Lu, Dongcai, Sun, Jiacheng, Xu, Qiang, Xin, Shen, Li, Zhenguo
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.07047
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909642459709440
author Xuejun, Yu
Zhong, Jianyuan
Feng, Zijin
Zhai, Pengyi
Yousefzadeh, Roozbeh
Ng, Wei Chong
Liu, Haoxiong
Shou, Ziyi
Xiong, Jing
Zhou, Yudong
Ong, Claudia Beth
Sugiarto, Austen Jeremy
Zhang, Yaoxi
Tai, Wai Ming
Cao, Huan
Lu, Dongcai
Sun, Jiacheng
Xu, Qiang
Xin, Shen
Li, Zhenguo
author_facet Xuejun, Yu
Zhong, Jianyuan
Feng, Zijin
Zhai, Pengyi
Yousefzadeh, Roozbeh
Ng, Wei Chong
Liu, Haoxiong
Shou, Ziyi
Xiong, Jing
Zhou, Yudong
Ong, Claudia Beth
Sugiarto, Austen Jeremy
Zhang, Yaoxi
Tai, Wai Ming
Cao, Huan
Lu, Dongcai
Sun, Jiacheng
Xu, Qiang
Xin, Shen
Li, Zhenguo
contents Recent advances in large language models show strong promise for formal reasoning. However, most LLM-based theorem provers have long been constrained by the need for expert-written formal statements as inputs, limiting their applicability to real-world problems expressed in natural language. We tackle this gap with Mathesis, the first end-to-end theorem proving pipeline processing informal problem statements. It contributes Mathesis-Autoformalizer, the first autoformalizer using reinforcement learning to enhance the formalization ability of natural language problems, aided by our novel LeanScorer framework for nuanced formalization quality assessment. It also proposes a Mathesis-Prover, which generates formal proofs from the formalized statements. To evaluate the real-world applicability of end-to-end formal theorem proving, we introduce Gaokao-Formal, a benchmark of 488 complex problems from China's national college entrance exam. Our approach is carefully designed, with a thorough study of each component. Experiments demonstrate Mathesis's effectiveness, with the autoformalizer outperforming the best baseline by 22% in pass-rate on Gaokao-Formal. The full system surpasses other model combinations, achieving 64% accuracy on MiniF2F with pass@32 and a state-of-the-art 18% on Gaokao-Formal.
format Preprint
id arxiv_https___arxiv_org_abs_2506_07047
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Mathesis: Towards Formal Theorem Proving from Natural Languages
Xuejun, Yu
Zhong, Jianyuan
Feng, Zijin
Zhai, Pengyi
Yousefzadeh, Roozbeh
Ng, Wei Chong
Liu, Haoxiong
Shou, Ziyi
Xiong, Jing
Zhou, Yudong
Ong, Claudia Beth
Sugiarto, Austen Jeremy
Zhang, Yaoxi
Tai, Wai Ming
Cao, Huan
Lu, Dongcai
Sun, Jiacheng
Xu, Qiang
Xin, Shen
Li, Zhenguo
Artificial Intelligence
Recent advances in large language models show strong promise for formal reasoning. However, most LLM-based theorem provers have long been constrained by the need for expert-written formal statements as inputs, limiting their applicability to real-world problems expressed in natural language. We tackle this gap with Mathesis, the first end-to-end theorem proving pipeline processing informal problem statements. It contributes Mathesis-Autoformalizer, the first autoformalizer using reinforcement learning to enhance the formalization ability of natural language problems, aided by our novel LeanScorer framework for nuanced formalization quality assessment. It also proposes a Mathesis-Prover, which generates formal proofs from the formalized statements. To evaluate the real-world applicability of end-to-end formal theorem proving, we introduce Gaokao-Formal, a benchmark of 488 complex problems from China's national college entrance exam. Our approach is carefully designed, with a thorough study of each component. Experiments demonstrate Mathesis's effectiveness, with the autoformalizer outperforming the best baseline by 22% in pass-rate on Gaokao-Formal. The full system surpasses other model combinations, achieving 64% accuracy on MiniF2F with pass@32 and a state-of-the-art 18% on Gaokao-Formal.
title Mathesis: Towards Formal Theorem Proving from Natural Languages
topic Artificial Intelligence
url https://arxiv.org/abs/2506.07047