Industry

ERC 评审两年前断言 LLM 无法胜任自动形式化

Why we picked it

Hugging Face 的 CTO 翻出两年前一份 ERC 评审报告,评审人当时一口咬定 LLM 做不了自动形式化。对比现在 LLM 在数学定理证明上的表现,这段旧评语读起来挺有意思,做学术申请或追踪 autoformalization 的人可以看看。

Julien Chaumond 在 X 上分享了一份 2024 年 ERC 申请的评审反馈截图。评审意见写道,该评审人不相信 LLM 能用于 autoformalization 及相关任务,因为其他团队已多次尝试且没有取得好的结果。

Body

这条推文以“Two years ago(两年前)”开头,暗指如今 LLM 在数学形式化上的进展已让这一判断过时。帖子由 Hugging Face 的首席技术官发出,引发对学术评审预判失误的讨论。

Original · Julien Chaumond

Two years ago 🤯 Marc Lelarge 🌻 @marc_lelarge Feedback from the Evaluation Report of an ERC application submitted in 2024: "This reviewer does not believe in the use of LLMs for autoformalization and related tasks: this has been tried several times by other groups without good results." 😭 🔗 View Quoted Tweet📊 4 ⚡

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Industry" desk, focusing on Europe's AI governance and technology ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include autoformalization, LLM, ERC, Hugging Face.

Source attribution: curated from Julien Chaumond (@julien_c).