Kimina-Prover Preview: Towards Large Formal Reasoning Models with Reinforcement Learning
Abstract
We present Kimina-Prover Preview, a state-of-the-art formal reasoning model developed using reinforcement learning techniques for Interactive Theorem Proving in Lean. Our approach demonstrates significant improvements on the miniF2F benchmark, showcasing the potential of RL-driven approaches for automated theorem proving.
Type
Publication
arXiv preprint arXiv:2504.11354
