ICLR authors say AI-reviewed peer review feels like 'pleasing an agent, not a person'

murat_kocaoglu_ · x · 2026-11-17

A PhD student vented about the ICLR rebuttal process: after days of unnecessary experiments answering odd reviewer demands (like never-heard-of "paired uncertainty" and questions already answered in the paper), they felt they were "no longer trying to convince a real researcher—just trying to please an agent," and that even a higher score would be approval from an agent, not a person. Murat Kocaoglu reposted with concern: established researchers giving up on peer review only works if you're already established, and the next generation suffers—calling for a review culture that values human judgment.

Original post →

More from Research

Research channel →