_version_ 1866915244095307776
author Wang, Haiming
Unsal, Mert
Lin, Xiaohan
Baksys, Mantas
Liu, Junqi
Santos, Marco Dos
Sung, Flood
Vinyes, Marina
Ying, Zhenzhe
Zhu, Zekai
Lu, Jianqiao
de Saxcé, Hugues
Bailey, Bolton
Song, Chendong
Xiao, Chenjun
Zhang, Dehao
Zhang, Ebony
Pu, Frederick
Zhu, Han
Liu, Jiawei
Bayer, Jonas
Michel, Julien
Yu, Longhui
Dreyfus-Schmidt, Léo
Tunstall, Lewis
Pagani, Luigi
Machado, Moreira
Bourigault, Pauline
Wang, Ran
Polu, Stanislas
Barroyer, Thibaut
Li, Wen-Ding
Niu, Yazhe
Fleureau, Yann
Hu, Yangyang
Yu, Zhouliang
Wang, Zihan
Yang, Zhilin
Liu, Zhengying
Li, Jia
author_facet Wang, Haiming
Unsal, Mert
Lin, Xiaohan
Baksys, Mantas
Liu, Junqi
Santos, Marco Dos
Sung, Flood
Vinyes, Marina
Ying, Zhenzhe
Zhu, Zekai
Lu, Jianqiao
de Saxcé, Hugues
Bailey, Bolton
Song, Chendong
Xiao, Chenjun
Zhang, Dehao
Zhang, Ebony
Pu, Frederick
Zhu, Han
Liu, Jiawei
Bayer, Jonas
Michel, Julien
Yu, Longhui
Dreyfus-Schmidt, Léo
Tunstall, Lewis
Pagani, Luigi
Machado, Moreira
Bourigault, Pauline
Wang, Ran
Polu, Stanislas
Barroyer, Thibaut
Li, Wen-Ding
Niu, Yazhe
Fleureau, Yann
Hu, Yangyang
Yu, Zhouliang
Wang, Zihan
Yang, Zhilin
Liu, Zhengying
Li, Jia
contents We introduce Kimina-Prover Preview, a large language model that pioneers a novel reasoning-driven exploration paradigm for formal theorem proving, as showcased in this preview release. Trained with a large-scale reinforcement learning pipeline from Qwen2.5-72B, Kimina-Prover demonstrates strong performance in Lean 4 proof generation by employing a structured reasoning pattern we term \textit{formal reasoning pattern}. This approach allows the model to emulate human problem-solving strategies in Lean, iteratively generating and refining proof steps. Kimina-Prover sets a new state-of-the-art on the miniF2F benchmark, reaching 80.7% with pass@8192. Beyond improved benchmark performance, our work yields several key insights: (1) Kimina-Prover exhibits high sample efficiency, delivering strong results even with minimal sampling (pass@1) and scaling effectively with computational budget, stemming from its unique reasoning pattern and RL training; (2) we demonstrate clear performance scaling with model size, a trend previously unobserved for neural theorem provers in formal mathematics; (3) the learned reasoning style, distinct from traditional search algorithms, shows potential to bridge the gap between formal verification and informal mathematical intuition. We open source distilled versions with 1.5B and 7B parameters of Kimina-Prover
format Preprint
id arxiv_https___arxiv_org_abs_2504_11354
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning
Wang, Haiming
Unsal, Mert
Lin, Xiaohan
Baksys, Mantas
Liu, Junqi
Santos, Marco Dos
Sung, Flood
Vinyes, Marina
Ying, Zhenzhe
Zhu, Zekai
Lu, Jianqiao
de Saxcé, Hugues
Bailey, Bolton
Song, Chendong
Xiao, Chenjun
Zhang, Dehao
Zhang, Ebony
Pu, Frederick
Zhu, Han
Liu, Jiawei
Bayer, Jonas
Michel, Julien
Yu, Longhui
Dreyfus-Schmidt, Léo
Tunstall, Lewis
Pagani, Luigi
Machado, Moreira
Bourigault, Pauline
Wang, Ran
Polu, Stanislas
Barroyer, Thibaut
Li, Wen-Ding
Niu, Yazhe
Fleureau, Yann
Hu, Yangyang
Yu, Zhouliang
Wang, Zihan
Yang, Zhilin
Liu, Zhengying
Li, Jia
Artificial Intelligence
We introduce Kimina-Prover Preview, a large language model that pioneers a novel reasoning-driven exploration paradigm for formal theorem proving, as showcased in this preview release. Trained with a large-scale reinforcement learning pipeline from Qwen2.5-72B, Kimina-Prover demonstrates strong performance in Lean 4 proof generation by employing a structured reasoning pattern we term \textit{formal reasoning pattern}. This approach allows the model to emulate human problem-solving strategies in Lean, iteratively generating and refining proof steps. Kimina-Prover sets a new state-of-the-art on the miniF2F benchmark, reaching 80.7% with pass@8192. Beyond improved benchmark performance, our work yields several key insights: (1) Kimina-Prover exhibits high sample efficiency, delivering strong results even with minimal sampling (pass@1) and scaling effectively with computational budget, stemming from its unique reasoning pattern and RL training; (2) we demonstrate clear performance scaling with model size, a trend previously unobserved for neural theorem provers in formal mathematics; (3) the learned reasoning style, distinct from traditional search algorithms, shows potential to bridge the gap between formal verification and informal mathematical intuition. We open source distilled versions with 1.5B and 7B parameters of Kimina-Prover
title Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning
topic Artificial Intelligence
url https://arxiv.org/abs/2504.11354