← 返回论文库
AlphaProof / AlphaGeometry 2: Olympiad-level Formal Mathematical Reasoning with RL
DeepMind · 2024
L1 · Domain Research AgentsNature 643 (2025)#formal-math#verification#mcts
CORE IDEA
AlphaProof(AlphaZero MCTS + Lean kernel)+ AlphaGeometry 2(LLM 提辅助线 + 符号引擎),IMO 2024 银牌。
CONCRETE EXAMPLE
AlphaProof 解决 IMO 2024 P6,现场只 5 人解出。
L-ANCHOR · 为什么在这一层重要
neuro-symbolic verification 终极形式,verifier 可信度 = Lean kernel
相关论文
Navigating the Alpha Jungle: An LLM-Powered MCTS Framework for Formulaic Factor Mining
FunSearch: Mathematical Discoveries from Program Search with LLMs
AI Scientist v1: Towards Fully Automated Open-Ended Scientific Discovery
AI Scientist v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search