AlphaProof methodology in Nature
Google DeepMind publishes AlphaProof's methodology in Nature. It uses reinforcement learning over formal Lean proofs and is the system behind the 2024 IMO silver.
Releases the technical detail of the reinforcement-learning system that proved IMO 2024 problems in Lean. It is the formal-proof counterpart to the natural-language Gemini Deep Think gold of 2025.
- Date
- Wednesday, 12 November 2025
- Lab
- Google DeepMind
- Kind
- paper
- Access
- research preview
Date and Nature publication come from the update note on the original 2024 post; the article itself was not opened.
Sources
This record was checked against its sources on 6 October 2026. How we check