Maxproof
arxiv.org
June 12, 2026
2 min read
🔥🔥🔥🔥🔥
53/100
Summary
MaxProof is a framework designed for population-level test-time scaling in mathematical proof, specifically within the MiniMax-M3 series. It trains three key capabilities: proof generation, proof verification, and critique-conditioned proof repair.
Key Takeaways
- MaxProof is a population-level test-time scaling framework designed for competition-level mathematical proof in the MiniMax-M3 series.
- The M3 model integrates proof generation, verification, and critique-conditioned proof repair capabilities, achieving scores of 35/42 on IMO 2025 and 36/42 on USAMO 2026.
- MaxProof employs a tournament selection method to refine and rank candidate proofs, ultimately producing a final proof.
- The M3 model's performance exceeds the human gold-medal threshold in mathematical competitions.
Coverage & discussions
Hacker News · arxiv.org
Community discussions
Community Sentiment
Positives
- The high fraction of gold medalists at the IMO suggests that training methods, potentially including AI tools, are effectively preparing students for complex problem-solving.
Concerns
- Concerns about the influence of LLMs on IMO preparation highlight potential ethical implications regarding fairness in competitive environments.
