Kimina-Prover

model paper
Moonshot AI
Formal reasoning model for Lean 4 theorem proving trained via RL from Qwen2.5-72B. Achieves 80.7% on miniF2F with pass@8192.
Paper (arXiv)GitHubHuggingFace

Outputs 2

Kimina-Prover-Preview-Distill-7B

model

Kimina-Prover Preview Paper

paper
arXiv HTML
reasoning