China has made a major breakthrough in the field of artificial intelligence and basic science. According to the Shanghai Artificial Intelligence Laboratory, the world-class mathematical problem that has plagued the global mathematics community for over 30 years—the three-dimensional Ghys conjecture—has been officially solved. This is also the first time that Chinese scientists have successfully tackled a world-class mathematical problem using domestic large models.
This research is led by the team of Xu Disheng, a visiting scholar at the Shanghai Artificial Intelligence Laboratory and a professor at Guangdong University in the Greater Bay Area. Relying on the “Shusen Science Multimodal Large Model S2” (Shusen-S2), the research team completed the proof of the three-dimensional Ghys conjecture through a human-machine collaboration approach based on the “integration of general and specialized technologies”. At the same time, Professor Dong Bin from Peking University and the team of Academician Liu Ruochuan used the SHEAF framework they developed independently to perform formal verification on the proof process, thereby clearly and accurately confirming the correctness of this research result.
As an important long-term open problem in the field of mathematical dynamical systems, the three-dimensional Ghys conjecture was proposed by French mathematician Étienne Ghys in 1995. The conjecture states that every holomorphic Anosov system on a compact complex manifold is biholomorphically conjugate to a holomorphic affine isomorphism on some complex infra-zero manifold. Due to its extremely high theoretical difficulty and symbolic importance in the field, it has long been recognized by the mathematical community as a touchstone for testing artificial intelligence and human mathematical reasoning abilities.
The Chinese scientific team successfully overcame this challenge through human-machine collaboration. This demonstrates the great potential of domestic large models in handling cutting-edge scientific computing and complex logical reasoning, and it also opens up a new practical path for deeply empowering basic scientific research with artificial intelligence in the future.