AI Model 'MaxProof' Achieves High Scores in Math Competitions
A new AI framework, MaxProof, has demonstrated advanced capabilities in generating and verifying mathematical proofs. Tested on simulated competition data, the system reportedly surpassed human gold-medal thresholds on the IMO 2025 and USAMO 2026 exams, highlighting potential advancements in AI for complex problem-solving.
Key points
- Researchers introduced MaxProof, a test-time scaling framework designed for generating and verifying mathematical proofs.
- The system integrates three core AI capabilities: proof generation, verification, and repair.
- When applied to the M3 AI model, MaxProof achieved scores of 35/42 on a simulated IMO 2025 and 36/42 on a simulated USAMO 2026.
- These simulated results exceed the known human gold-medal thresholds for both competitions.
- The framework utilizes a population-based search and tournament selection to refine proof candidates.
A novel AI framework named MaxProof is making waves in the field of artificial intelligence for its performance on complex mathematical reasoning tasks. Developed by a team of researchers, MaxProof is designed to enhance the capabilities of AI models in generating and verifying mathematical proofs.
The system operates by training an AI model, dubbed M3, on three key proof-oriented skills: generating proofs, verifying their correctness, and repairing flawed ones. This integrated approach aims for a robust and accurate proof-solving mechanism. MaxProof then employs a population-level testing strategy, treating the AI model as a generator, verifier, refiner, and ranker.
During evaluation, the MaxProof framework was applied to the M3 model using data simulating the International Mathematical Olympiad (IMO) 2025 and the American Invitational Mathematics Examination (AIME) 2026. The results, as presented in a pre-print paper, indicate that the AI achieved scores of 35 out of 42 on the simulated IMO 2025 and 36 out of 42 on the simulated USAMO 2026. The researchers state these scores surpass the typical human gold-medal benchmark for these prestigious competitions, suggesting a significant leap in AI's potential for high-level academic problem-solving.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.