CHINA / SOCIETY
China-developed AI makes history as its geometry questions enter top human math contests
Published: Jan 30, 2026 11:50 PM
Photo: VCG

Photo: VCG

A universal artificial intelligence (AI) system developed by a Chinese research team has become the world's first to solve nearly all Olympic-level geometry questions from the past 25 years within a maximum of 38 minutes, with geometry questions generated by the system now included in prestigious human mathematics competitions, Science Daily reported.

The research team led by Zhang Chi at the Beijing Institution for General Artificial Intelligence has developed TongGeometry, the world's first general AI system endowed with dual capabilities of autonomous problem proposing and automatic problem solving. 

The research findings were published in Nature Machine Intelligence on January 26.

Zhang explained that TongGeometry can pinpoint and extract high-quality problems that meet the aesthetic benchmarks of human mathematicians from an immense corpus of spatial configurations.

This marks the world's first paradigm shift in the field from "imitative problem solving" to "autonomous creation," the report said.

In contrast to AlphaGeometry, an AI Olympiad-level AI system for geometry developed by DeepMind in early 2024, TongGeometry demonstrates a higher level of intelligence. 

It is not merely a top performer that aces every assessment with a perfect score, but also a skilled problem composer capable of creating elegant and original mathematical problems, Science Daily reported.

TongGeometry's original problem-creating capability has already been recognized by both academic and competition circles, the report said.

Three original geometry questions autonomously generated by the system have been officially included in the 2024 national high school mathematics competition (Beijing Division) and the US Ersatz Math Olympiad, marking the first time AI-originated problems have been admitted to high-caliber human mathematics competitions.

This achievement signifies that the Chinese research team has attained key technological self-reliance in the core domain of automated reasoning, with performance and functional diversity surpassing top international benchmarks represented by DeepMind, according to the report.


Global Times