🏆 LEAP: AI Solves All Putnam 2025 Olympiad Problems
Google has introduced LEAP (LLM-in-Lean Environment Agentic Prover) — an agentic framework for automated theorem proving in the Lean language. The system utilizes general-purpose LLMs, breaking tasks down into subtasks and using the Lean compiler to correct errors. LEAP solved all 12 problems of the Putnam 2025, surpassing Gemini 3.1 Pro.
🌍 The transition to agentic systems paves the way for scalable formal mathematical reasoning.
👤 This is a step toward creating AI with absolutely reliable reasoning, verifiable by mathematical software.
Source 1: https://arxiv.org/pdf/2606.03303 Source 2: https://github.com/google-deepmind/superhuman/tree/main/leap/solutions/Putnam-2025
