Researchers at Peking University have developed an artificial intelligence system capable of solving a mathematical conjecture that has remained unsolved for over a decade.
The system successfully formalized the solution to a problem in commutative algebra originally proposed by American mathematician Dan Anderson in 2014.
According to the South China Morning Post, the AI completed the solution in just a few hours, with the full verification process taking approximately 80 hours of execution time.
Dual-agent architecture
The breakthrough relies on a specialized "double agent" model designed to bridge the gap between human intuition and machine logic.
One agent performs informal reasoning to explore various strategies and construct potential proofs using natural language. A second agent then translates these ideas into a rigorous mathematical format that a machine can verify with absolute certainty.
This process automates the transition from hypothesis to formal proof, a task that traditionally requires years of collaboration between human specialists.
The development team noted that human involvement was limited to providing the system with access to restricted documents that the AI could not retrieve independently. No human intervention was required to guide the mathematical logic of the solution itself.
While the results are promising, the authors noted in a preliminary article on the arXiv repository that the findings have not yet undergone peer review.
Reliability remains a primary hurdle for the field, as researchers struggle to validate AI-generated proofs without independent external verification. The team suggests that combining natural language reasoning with formal verification could eventually increase confidence in automated mathematical discoveries.
This achievement highlights China's expanding influence in the global AI race, following recent high-profile developments from companies like Alibaba, ByteDance, and DeepSeek.
