Ax-Prover: Multi-Agent System for Automated Theorem Proving in Math and Physics
Ax-Prover is a multi-agent system for automated theorem proving in Lean, capable of solving problems across scientific domains either autonomously or with human collaboration. It integrates Large Language Models (LLMs) for reasoning with Lean tools via the Model Context Protocol (MCP) to ensure formal correctness. Benchmarked against frontier LLMs and specialized provers on public math datasets and new Lean benchmarks in abstract algebra and quantum theory, Ax-Prover achieves competitive performance and outperforms existing systems.
Key facts
- Ax-Prover is a multi-agent system for automated theorem proving in Lean.
- It operates autonomously or collaboratively with human experts.
- It uses LLMs for reasoning and Lean tools via MCP for formal correctness.
- Benchmarked on public math benchmarks and new Lean benchmarks in abstract algebra and quantum theory.
- Competitive with state-of-the-art provers on public datasets.
- Largely outperforms existing systems on introduced benchmarks.
- Published on arXiv with ID 2510.12787.
- Covers mathematics and quantum physics domains.
Entities
Institutions
- arXiv
- Lean