A math AI research alliance led by Fields Medalist Shing-Tung Yau (丘成桐) launched at Tsinghua University in Beijing on October 11 Eight Chinese mathematics research institutions are taking part, and it said it expects about 1,000 researchers to participate

In June of the same year, mathematicians released the Leiden Declaration, calling for verification of AI proofs and disclosure of tools The two moves differ in how they approach AI math research

Concept shot of a blackboard with no equations and a network of connected institutes Concept shot of a blackboard with no equations and a network of connected institutes — Source: concept shot, self-generated with agy

Makeup of the alliance

Its official name is the Yau Laboratory Alliance for Mathematics and AI The participating institutions reported by Science and Technology Daily (科技日报) on October 13 are the eight below

  • Qiuzhen College (求真书院), Tsinghua University
  • Yau Mathematical Sciences Center, Tsinghua University
  • Beijing Institute of Mathematical Sciences and Applications (BIMSA)
  • Morningside Center of Mathematics, Chinese Academy of Sciences
  • Shanghai Institute for Mathematics and Interdisciplinary Sciences
  • Yau Center, Southeast University (Nanjing)
  • Hetao Institute of Mathematics and Interdisciplinary Sciences (Shenzhen)
  • Chih-Chen College, The Chinese University of Hong Kong

Yau heads the alliance, and academic leads nominated by each institution set research directions and evaluation criteria in a committee Participating institutions share large-scale model training resources, formalized mathematics libraries, and computing platforms

A plan to set up math AI institutes in five places, Beijing, Shanghai, Nanjing, Shenzhen, and Hong Kong, was announced as well

Goals and current stage

The goals are math-specialized AI models, AI research infrastructure, interdisciplinary research, and talent development VnExpress and others also reported proposing new mathematical ideas and generating proofs researchers can verify as goals

No models or results have been released yet The 1,000 figure was also announced not as current headcount but as the number expected to participate in the future

🗣️ Yau’s diagnosis

Yau said Chinese AI models do everyday tasks well But he said they lag far behind leading US systems in advanced mathematics and physics

His explanation is that US companies mobilize thousands of researchers while China’s research workforce is scattered across many organizations

According to SCMP, he said that a year ago he didn’t know AI would be this powerful He compared AI’s progress in mathematics to a tsunami

The Leiden Declaration’s principles for AI verification

Leiden Declaration homepage with the International Mathematical Union (IMU) endorsement Leiden Declaration homepage with the International Mathematical Union (IMU) endorsement — Source: Leiden Declaration

The Leiden Declaration was released on June 2, 2026, and endorsed by the International Mathematical Union (IMU) Sixteen researchers who gathered at a workshop at the Lorentz Center in Leiden, the Netherlands, in September 2025 wrote the draft

According to the declaration’s site, there are 4,357 signatories as of October 19 It doesn’t call for banning AI use, but asks for the following

Principle Content
Verification Automatically generated proofs are subject to independent verification
Disclosure State the AI tools used in research
Attribution If attribution isn’t possible, say so explicitly

IMU Vice President Ulrike Tillmann said mathematics should remain a deeply human activity

Key 2026 dates in math and AI

2026 timeline of the OpenAI disproof, the Leiden Declaration, OpenAI's results release, and the launch of Yau's alliance 2026 timeline of the OpenAI disproof, the Leiden Declaration, OpenAI’s results release, and the launch of Yau’s alliance — Source: self-rendered based on OpenAI, Leiden Declaration, Science and Technology Daily

On May 20, OpenAI announced that an internal model had disproved Erdős’s planar unit distance conjecture The Leiden Declaration came right after, on June 2

On October 6, OpenAI released more math results from its internal models Some proofs were formalized in Lean and posted alongside

I covered OpenAI’s proof of the unique games conjecture in an October 13 post

🔗 Link - https://blog.naver.com/rottensweetpotato/224440548862

How Claude formally verified Fermat’s Last Theorem in Lean is in a September post

🔗 Link - https://blog.naver.com/rottensweetpotato/224411365394

What to watch next

  • First output: when Yau’s alliance releases a math-specialized model
  • Verification: whether the alliance adopts formal verification such as Lean as a standard
  • Leiden Declaration: whether journals introduce AI-use disclosure rules
  • Korea: a public response from the Korean math community is not confirmed as of October 19

Sources