AI Achieves Silver Medal Standard in Solving IMO Problems with AlphaProof and AlphaGeometry 2

AI Achieves Silver Medal Standard in Solving IMO Problems with AlphaProof and AlphaGeometry 2

On July 25, 2024, DeepMind announced a significant breakthrough in artificial intelligence (AI) with the development of two models: AlphaProof and AlphaGeometry 2. These systems have achieved a silver medal standard by solving four out of six problems from this year’s International Mathematical Olympiad (IMO). This marks a pivotal moment for AI in advanced mathematical reasoning.

Technical Breakthroughs

AlphaProof

• Reinforcement Learning: AlphaProof is built on reinforcement learning, a type of machine learning where the model learns to make decisions by trial and error. It receives rewards or penalties based on its performance.
• Formal Math Reasoning: The system excels in formal math reasoning, which involves proving mathematical statements with rigorous logical steps. This capability is crucial for solving complex problems that require deep understanding and precise logic.
• Training Data: AlphaProof was trained on a vast dataset of mathematical proofs and problems, including historical IMO questions. This extensive training helps the model generalize to new, unseen problems.

AlphaGeometry 2

• Improved Geometry-Solving System: An enhanced version of DeepMind’s previous geometry-solving system, AlphaGeometry 2 has been optimized for better performance in geometric problem-solving.
• Advanced Geometric Algorithms: The system uses advanced algorithms to handle complex geometric configurations and relationships. This includes techniques like computational geometry and symbolic computation.
• Scalability and Efficiency: AlphaGeometry 2 is designed to be more scalable and efficient, allowing it to solve problems faster and with greater accuracy.

Performance at the IMO

The International Mathematical Olympiad (IMO) is a prestigious competition for pre-college mathematicians, held annually since 1959. Each year, participants tackle six exceptionally difficult problems in algebra, combinatorics, geometry, and number theory. The problems are designed to test not only mathematical knowledge but also creative problem-solving skills.

This year, DeepMind’s combined AI system was applied to the IMO competition problems, provided by the organizers. The solutions were evaluated according to the IMO’s scoring rules by prominent mathematicians, including Prof Sir Timothy Gowers (an IMO gold medalist and Fields Medal winner) and Dr Joseph Myers (a two-time IMO gold medalist and Chair of the IMO).

• Solved Four Out of Six Problems: AlphaProof and AlphaGeometry 2 managed to solve four out of the six problems, achieving a performance equivalent to a silver medal in the competition.
• First AI System to Reach This Level: This is the first time an AI system has reached this level of performance at the IMO, demonstrating significant progress in advanced mathematical reasoning.

Implications for AI and Mathematics

The success of AlphaProof and AlphaGeometry 2 highlights the potential of AI in advancing mathematical research and education. Here are some key implications:

• Enhanced Problem-Solving Capabilities: These models can assist mathematicians in discovering new insights, algorithms, and solutions to open problems.
• Educational Tools: The systems could be used as educational tools to help students develop their problem-solving skills and deepen their understanding of advanced mathematical concepts.
• Research Applications: AI with advanced reasoning capabilities can contribute to various scientific and technological fields, from cryptography to physics.

Future Directions

DeepMind’s achievement is a significant step forward in the development of artificial general intelligence (AGI) with advanced mathematical reasoning. The team plans to continue refining these models and exploring new applications. On November 12, 2025, they published the methodology behind AlphaProof in an article in Nature, providing further insights into their approach.

As AI continues to evolve, we can expect more breakthroughs in solving complex problems across various domains. The success of AlphaProof and AlphaGeometry 2 is a promising sign of what the future holds for AI in mathematics and beyond.