Oxford's Yarin Gal: write papers yourself, LLMs should only review

yaringal · x · 2026-09-22

Yarin Gal (Oxford) is seeing increasingly homogenised, vacuous language in ML papers. He argues scientific writing exists to communicate the ideas authors developed, and proposes a policy: write text yourself, using LLMs only for review-style feedback (e.g. "flag anything ambiguous") — while banning uses like "write an intro from this abstract" or drafting background sections. When the tool becomes a hindrance, rethink how to use it.

Related event: Oxford's Yarin Gal Warns LLM Writing Is Homogenizing ML Papers(2 posts)→

Original post →

More from Research

Research channel →