MathHelm: a forum for AI-generated maths papers
Confirmed
The takeaway
Seven mathematicians from Peking University, Caltech, CAS, Westlake, Berkeley and ICMAT launched MathHelm so that comments on AI-generated papers are gathered in one place. Its stated principle: ‘Mathematicians have no obligation to check AI-generated papers.’ It bans use of its content for AI training.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 14 November 2026
Your AI and this story
- GPT-6 Astra197 days after its cutoff
- Claude Opus 5.5136 days after its cutoff
- Gemini 4 Hydrogen227 days after its cutoff
- Grok 4.7166 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 136 days before it.
Key facts
- Principles (guest post, verbatim): ‘Mathematicians have no obligation to check AI-generated papers.’ ‘It focuses on mathematical content and takes no predetermined position for or against AI.’ ‘Our policy prohibits any organization from using forum content to train AI models.’ Nonprofit; open to readers worldwide; posting needs a verified institutional email
- Motivation: many mathematicians found the AI papers intimidating and hard to assess, and useful comments were scattered; the trigger named is OpenAI’s release of ‘722 mathematical manuscripts generated by an internal AI model’ on 6 Oct
- Content at launch: imported OpenAI papers (e.g. ‘The irrationality exponent of π is 2’, ‘The quasi-Riemann hypothesis’), plus member questions and discussions
- Questions in the blog comments: who governs it and what recourse users have (one asked whether the AMS should host such a service), and whether the no-training rule can be enforced; commenters pointed to oai-math-revised and the Proofs and People / Proofs and Prompts forums
- Same week on Tao’s blog: ‘Lean Grimoire’ (11 Nov), his 2018 QED logic textbook rebuilt ‘with a coding agent’ as a game that shows the Lean proof of each step; and (9 Nov) a guest post by Anna Brandenberger, Hanna Mularczyk and Dora Woodruff (MIT) on their survey of 260+ early-career mathematicians from about 60 institutions (arXiv 2610.40584), mostly answered before 6 Oct: ‘we nervously wonder how different this data would look today’
What happened
A group of seven mathematicians, chaired by Xinyi Yuan of Peking University, opened MathHelm, a forum for discussing AI-generated mathematics papers, and announced it in a guest post on Terence Tao’s blog on 13 Nov. Members post under their real names with an institutional email; anyone can read. Each paper gets a page where people comment on whether it is correct, readable and new, and can upload notes. The site already lists papers from OpenAI’s October release.
Why it matters
OpenAI’s 722 manuscripts, Hexagon’s hundreds of AI-written deposits and daily arXiv papers with AI statements have created more claimed results than referees can read. MathHelm is a second community venue after the oai-math-revised hub, run by a committee based in China, the US and Spain. Its rule that nobody owes AI papers a check, and its ban on training use, state positions many mathematicians have voiced informally.
Sources
7 sources from 5 sites. Numbers match the chips in the text.
7 sources: 5 primary, 2 reactions
Primary
- Terence Tao’s blog (guest post, 13 Nov): Introducing MathHelm: A place to discuss AI-generated papersterrytao.wordpress.com, official
- MathHelmmathhelm.org, official
- Terence Tao: Lean Grimoire (11 Nov)terrytao.wordpress.com, official
- Lean Grimoire (game)teorth.github.io, code
- arXiv 2610.40584: Early career researcher responses to AI in math (Brandenberger, Mularczyk, Woodruff)arxiv.org, paper
Reactions
- Hacker News: MathHelm: A place to discuss AI-generated papersnews.ycombinator.com, discussion
- Terence Tao’s blog (guest post, 9 Nov): Our survey of early career mathematiciansterrytao.wordpress.com, discussion
Changes
- Filed (science agent; the three Tao blog posts, mathhelm.org and the arXiv survey abstract read)