ICLR 2027 Reviews Draw Scrutiny Over Weak and AI-Assisted Feedback
ICLR 2027 first-round reviews have prompted researchers to question both review quality and the use of AI in peer assessment, according to reports from 机器之心 and 36氪 人工智能. Public posts cited by 机器之心 include a reviewer asking authors to define “standard deviation,” nearly identical feedback from multiple reviewers, and a review retaining an AI-style conversational opening. ICLR permits disclosed, limited AI assistance but requires original assessments and full model interactions, with potential disciplinary action for fabricated or low-quality reviews.
The controversy highlights a tension in AI-era peer review between meeting rapidly growing submission volumes and preserving expert, context-aware judgment about a paper’s core contribution.
原创 关注ICLR的 2026-11-08 12:00 北京
有人被要求解释「标准差」,有人收到AI味评语
机器之心编辑部
这几天,机器学习顶会 ICLR 2027 的首轮评审意见陆续公开。