#Autoformalization
A central goal of the AI-for-math community has been to figure out how to use AI systems to translate mathematical arguments into formally verifiable code, called “autoformalization.” Like so much else, AI autoformalization capabilities changed this summer. www.quantamagazine.org/updates/tran...
Transformation | Quanta Magazine
A weekly dispatch exploring the latest news in math and AI…
www.quantamagazine.org
November 16, 2026 at 2:03 PM
Last week, an AI startup announced four autoformalization results, which independent experts say they are satisfied with. It now seems a question not of if, but when, all of human mathematical knowledge will be formalized by AI.

Read more on Transformation: www.quantamagazine.org/updates/tran...
Transformation | Quanta Magazine
A weekly dispatch exploring the latest news in math and AI…
www.quantamagazine.org
November 14, 2026 at 2:01 PM
What is autoformalization in AI and math?

Autoformalization is when an AI automatically translates math written in human language (ordinary sentences and formulas), such as a problem or a proof, into a formal language (one in which the meaning of every symbol is fixed exactly, so a computer can…
What is autoformalization in AI and math?
Autoformalization is when an AI automatically translates math written in human language, such as a problem or a proof, into a proof language that a computer can check, like Lean or Isabelle. Here is why math is translated, how accurate AI translations are, the trap of a wrong translation and how to catch it, and how it is used in systems such as AlphaProof, explained with pictures and a quick quiz.
naru.news
November 10, 2026 at 9:33 PM
自動形式化(AIが数学を証明の言葉に訳すこと)とは?

自動形式化とは、人の言葉(ふつうの文と数式)で書かれた数学の問題や証明を、AIが、形式の言葉(記号1つ1つの意味がきっちり決めてあり、コンピューターが1行ずつ正しさを確かめられる言葉)に、自動で訳すことです。
自動形式化(AIが数学を証明の言葉に訳すこと)とは?
自動形式化(Autoformalization)とは、人の言葉で書いた数学の問題や証明を、AIが、Lean・Isabelle などの、コンピューターが正しさを確かめられる証明の言葉に、自動で訳すこと。なぜ訳すのか、AIの訳はどのくらい正しいか、訳のまちがいという落とし穴と見つけ方、AlphaProof などでの使われ方を、図と確認問題で、世界一わかりやすく解説します。
naru.news
November 10, 2026 at 9:33 PM
PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean
Read more: https://arxiv.org/html/2610.31810v2
November 3, 2026 at 3:42 AM
I'm also recruiting through this postdoc call -- anything related to my interests + AI, but please especially reach out of you're interested in AI autoformalization in Lean.

(To see some of my work here, see gargnikhil.com/AppliedModel... and this preprint: arxiv.org/abs/2606.13306)
November 1, 2026 at 9:52 PM
"The… preprint also includes a proof that autoformalization is infinitely harder than any other computational problem. The reason boils down to the ambiguous, sloplike nature of the natural language proofs generated by AI…
AI is shaking up math. But can its proofs be trusted?
As AI firms churn out math breakthroughs, researchers question use of Lean software as proof checker
www.science.org
November 1, 2026 at 7:30 PM
Always Be Creating grant proposals that the NSF approves.
October 29, 2026 at 5:28 PM
Leo de Moura, creator of Lean, discusses how AI-driven autoformalization is revolutionizing software verification and mathematics. By making machine-checked proofs practical, AI enables the creation of verified software stacks, transforming how we ensure reliability.
AI, Lean, and the Future of Formal Verification, with Lean Creator Leo de Moura (104)
news.ycombinator.com
October 31, 2026 at 2:56 PM
Breaking the chains of human-written policies, we've unleashed an autoformalization pipeline.

Translating agent instructions, tool descriptions, and natural language policies into formally verified policies, it's a game-changer.
October 29, 2026 at 5:31 AM
“If the future of Lean is that every interaction is mediated through autoformalization, for me personally it would be a very dull future.”

I thought this was quite an interesting article. #MathSky #LLM

www.science.org/content/arti...
AI is shaking up math. But can its proofs be trusted?
As AI firms churn out math breakthroughs, researchers question use of Lean software as proof checker
www.science.org
October 27, 2026 at 10:02 PM
Half agree and half disagree. Agree on the sadness of autoformalization sucking up oxygen. Disagree on the extent of doubt this article casts on lean itself. Like any formal guarantee what lean does is define and limit the surface area of proofs. Also the dunk on type theory makes no sense to me.
October 27, 2026 at 10:23 PM
PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean

Jiyuan Tan, Vasilis Syrgkanis, Jose Blanchet

#arXiv #cs.AI #cs.LO
PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean
Even a carefully proofread paper can leave small gaps in its proofs or its stated assumptions. We introduce PaperFormalizer, a Lean-based agentic pipeline that helps authors find and close such gaps in their own papers and turn each paper into a Lean-verified version. Starting from the LaTeX source…
arxiv.org
October 28, 2026 at 2:01 AM
PaperFormalizer: Author-in-the-loop autoformalization of research papers in Lean. ~ Jiyuan Tan, Vasilis Syrgkanis, Jose Blanchet.

arxiv.org/abs/2610.31810

#LeanProver #ITP #AI4Math #Autoformalization
PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean
Even a carefully proofread paper can leave small gaps in its proofs or its stated assumptions. We introduce PaperFormalizer, a Lean-based agentic pipeline that helps authors find and close such gaps i...
arxiv.org
October 26, 2026 at 11:41 AM
Interesting account of autoformalization via open-weights models on a desktop: www.mctague.org/carl/blog/20...
Autoformalization on a PC, Carl McTague's Blog
www.mctague.org
October 25, 2026 at 3:21 AM
Jiyuan Tan, Vasilis Syrgkanis, Jose Blanchet: PaperFormalizer: Author-in-the-Loop Autoformalization of Research Papers in Lean https://arxiv.org/abs/2610.31810 https://arxiv.org/pdf/2610.31810 https://arxiv.org/html/2610.31810
October 26, 2026 at 6:38 AM
Harvard researchers present a way to verify Lean formalizations using LLM checks and numerical and vacuity checks. This reduces error rates and reveals dataset flaws, enhancing autoformalization via an Extract-Verify-Prove pipeline for reliable math libraries. https://arxiv.org/abs/2610.20573
Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
ArXiv link for Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
arxiv.org
October 24, 2026 at 9:10 AM
Yeah they are gluttons for this sort of analytical autoformalization attempts and they are good at solving math problems but the philosophical payoff is uncertain
October 22, 2026 at 10:30 AM
A lot of this just looks like if I want to autoformalization how I think about meta theory making how would I do it
October 18, 2026 at 9:29 PM
IUT正しいと思うわ。

数学者が大したことないことがよーくわかった。望月さんが賢すぎて理解出来ないんでしょ

x.com/IUTTOfSM1969...
IUTGtr (@IUTTOfSM1969) on X
IUT理論を構築した著者が研究チームを組織、(AIによるautoformalizationではなく)Lean専門家とのコミュニケーションを通じて誤解の核心となる部分(→Stages 1~2)に対するコードの骨格を作成。疑われている自然言語ベースの論証の間のgapが存在しないことを示すことで、理論の正否に終止符を打つ。
x.com
October 19, 2026 at 1:54 PM
that maybe your field could have worked harder not smarter(though AI still fails when no formal verification allows the brute force approach to work, and even then it can fail, say autoformalization in math). But as we all know if AI really gets good most people will be fired, not given personal
October 18, 2026 at 9:48 PM
A verification method for Lean formalizations combines LLM reviews with checks, achieving the lowest error rate in formalizing functions. This method addresses reliability issues in autoformalization, paving a trusted path for advanced mathematical research. https://arxiv.org/abs/2610.20573
Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
ArXiv link for Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
arxiv.org
October 16, 2026 at 3:10 PM
Harvard researchers created a verification method improving Lean formalizations by merging LLM reviews with numerical and vacuity checks. This approach lowers error rates and uncovers translation defects, enabling more reliable automated mathematical formalization. https://arxiv.org/abs/2610.20573
Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
ArXiv link for Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization
arxiv.org
October 16, 2026 at 12:10 PM