Research
114d agoAI Peer Reviewers Are Easily Gamed
LLM-generated reviews for ICLR 2026 exhibit a hivemind effect that erases perspective diversity. Researchers found that simply prompting an AI to rewrite a paper—a process called paper laundering—significantly inflates review scores. This vulnerability proves that current models lack the rigor for academic gatekeeping.
The Signal
Practitioners should avoid automating peer review without new evaluation frameworks.