Oxford's Yarin Gal Proposes arXiv Ban LLM-Written Papers to Curb AI Slop

yaringal · x · 2026-09-23

Yarin Gal argues arXiv should forbid LLM-written papers outright, rejecting flagged submissions via watermarking: authors may use LLMs to improve language and reviews, but not to write the papers. Filtering writing only—not LLM-assisted experiments or proofs—would massively reduce slop by making spam far harder, with readers applying critical scrutiny to LLM-generated proofs as before.

Related event: arXiv Responds to LLM Paper Controversy: Detect but Don't Block(2 posts)→

Original post →

More from Safety

Safety channel →