2026 · News A new direction for AlphaScience Lab

Researcher · Tokyo

AI for Science,
across disciplines.

Research in AI-assisted science and mathematics, together with writing about how that research is done.

AlphaScience Lab · Research and writing

Research, methods,
and the work in progress.

AI for Science spans more than one method or one paper. Here are selected preprints, research practice, and writing. Each work is labeled by its current publication and verification status.

01 / Selected work

Research with a record.

AI for Science · arXiv preprint

Where Scientific Search Agents Fail: Decision-Checkpoint Auditing of Exposure and Inspection Attempts

Research by Hongmin Li and Wanli Zhao on how to audit exposure to and attempts to inspect target papers during scientific search.

Read paper ↗
Mathematics · ResearchGate preprint

The Hodge Conjecture for Odd-Degree Fermat Fourfolds

Preprint by Hongmin Li. Claims and proofs require independent verification.

View preprint ↗
Mathematics · ResearchGate preprint

Low-t Coefficients of the Valley Delta Conjecture

Preprint by Hongmin Li. Claims and proofs require independent verification.

View preprint ↗
Mathematics · Preprint

Resolutions of linear p-cyclic quotient singularities

Research by Linghu Fan and Hongmin Li. Read the paper for its results and scope.

Read paper ↗
Methodology · Preprint

The Calibration Turn in AI-Assisted Research

A conceptual framework for matching scientific claims to the evidence available.

Read paper ↗
02 / Writing

From research to conversation.

Blog · Medium

A Mathematician, a Non-Mathematician, and a Machine

A first-person account of what we learned running AI on a real mathematical research problem.

Read the essay ↗

03 / Connect

Research is a conversation.

Interested in AI for Science, mathematical research, or how research claims are evaluated? Explore the work or get in touch.