Researchers from the Greater Bay Area University and the Shanghai AI Lab have solved a 30-year-old mathematical problem. The team used the locally developed Intern-S2 large language model to prove a hypothesis proposed by French mathematician Étienne Ghys in 1995 regarding chaotic dynamic systems.
The project involved a collaborative approach where the AI performed logical reasoning, literature analysis, and iterative searching, while human scientists provided guidance and structural oversight. Formal verification of the proof was conducted by two teams from Peking University using the SHEAF system.
The Shanghai AI Lab intends to integrate this workflow into a reusable mathematical verification module for its research platform. This platform, which combines the Intern-S2 model with AI agents and experimental hardware, has already reduced development cycles in electrochemical synthesis from 5–10 years to 1–2 years.
The researchers emphasize that this methodology is applicable to fields such as physics, chemistry, and biology. The AI acts as a research assistant, handling routine tasks like verification calculations and hypothesis decomposition to accelerate scientific discovery.
