ICML 2026 desk rejected 497 papers, approximately 2% of all submissions, because reviewers used large language models (LLMs) in violation of their explicit agreements not to do so. This action targeted submissions reviewed under Policy A, which strictly prohibits LLM use. The program chairs and scientific integrity committee identified 506 reciprocal reviewers who breached their commitments, leading to the rejections. Five hundred and ninety-five reviews (~1% of all reviews) written by these reviewers were flagged for LLM use and removed.
If a Policy A reviewer generated more than half their reviews with an LLM, all their reviews were deleted and they were removed from the reviewer pool. Ten percent of the 506 detected violators fell into this category.
Source: Hacker News · Summarized by HeadlinesBriefing