Large-Scale Study Finds LLMs Measurably Shifting Peer Review Opinion Patterns at Top AI Conferences

arXiv·medium signal

Wu and Zhang analyze the impact of LLMs on peer review opinions at top AI conferences from a fine-grained perspective, finding measurable shifts in how reviews are written post-LLM availability. The study examines whether LLM-generated or LLM-assisted reviews differ systematically from fully human reviews across dimensions like specificity, constructiveness, and recommendation patterns. Raises questions about the integrity of the review process that governs which AI research gets published.