TLDRocket
Sign in

Kimina-Prover-RL

Hugging Face

Kimina Prover released kimina-prover-rl, an open-source Lean 4 formal theorem proving reinforcement-learning training pipeline compatible with the Verl framework. The 1.7B model AI-MO/Kimina-Prover-RL-1.7B reached 76.63% Pass@32 on the MiniF2F benchmark. The release adds verifiable two-stage reasoning-then-code training with format checking and one error-correction turn, and it comes with a reproducible training recipe and two newly released models.

Read more related to this at Hugging Face
Kimina Prover kimina-prover-rl AI-MO/Kimina-Prover-RL-1.7B Verl Lean 4 MiniF2F

Related stories

Kimina-Prover: Applying Test-time RL Search on Large Formal Reasoning Models Hugging Face · 1 year ago · 38 Kimi AI and kvcache-ai Open Sources ‘AgentENV’: A Distributed System that Powers Agentic Reinforcement Learning (RL) Training for Kimi K3 MarkTechPost · 3 months ago · 18
Kimi K3: The Complete Developer Guide Together AI · 2 months ago · 58
Open Source Models Reinforcement Learning

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.