What’s New in ICML 2026 Peer Review
By ICML 2026 Program Chairs Alekh Agarwal, Miroslav Dudik, Sharon Li, Martin Jaggi.
As ICML 2026 opens for submissions, we would like to review several new elements that we are introducing this year to improve the reviewing process and to adapt to the changing landscape of LLM use in research.
Thinly sliced contributions
There is a rising concern in our community about a large volume of submissions that do not necessarily advance the field, while burdening the peer-review system. The scale of this issue is exacerbated by the use of generative AI tools. We are taking aim at what we call “thinly sliced contributions”, by which we mean submission of a large number of papers with small variations on the same theme.
To combat this phenomenon, we are strengthening our concurrent submission policy. The policy states that all papers submitted to ICML 2026 by the same author must be viewed as prior work, and if they are on a related topic (which would be normally included in the related work section of the paper), they must cite each other and discuss each other in the the body of the paper (anonymized PDFs of the cited submissions must be provided in the supplementary material). Violations of this policy are viewed as abuse of the peer-review system and may result in the desk rejection of the violating submissions or of all the submissions by the submitting authors.
Cascading desk rejections for peer-review abuse
Violations of integrity of peer-review process through prompt injection, collusion, as well as other forms of abuse like submission of low-quality AI-generated content (AI slop) and submission of thinly sliced contributions may result in penalties as outlined in Peer-review Ethics. In particular, these violations may result in desk rejection of all papers by any of the authors of an offending paper. This places an additional responsibility on group leaders. If group leaders are co-authors on papers that abuse the peer-review system, all of the papers where they are listed as an author might be desk rejected regardless of whether they were aware of the specific instance of peer-review abuse. We urge group leaders to be vigilant about the papers they submit and work closely with all of the contributing authors to avoid cascading desk rejections that would impact all work they are submitting, and all authors they are submitting with.
Reciprocity
As in the past years, we expect paper authors to give back to the community for the service that they receive by having their papers reviewed by their peers. As such, each paper must nominate a qualified reciprocal reviewer. Authors that submit four or more papers must be involved as a reviewer, area chair, senior area chair or in other organizing roles at ICML (details of this policy, including exceptions, are spelled out in the Call for Papers). Moreover, if any reviewer or area chair fails in their duties, all of their papers might be desk rejected (details in Peer-review Ethics).
Supporting authors and reviewers
The policies discussed so far seek to improve accountability of authors and reviewers. We also have several policies that seek to meet various author and reviewer needs:
- Our flexible policy for LLMs in reviewing allows a broad (but not completely unconstrained) use of LLMs in reviewing, as long as authors consent to that form of review.
- Top 25% of reviewers will receive free ICML registration.
- We are integrating author-provided self-rankings (for authors with multiple submissions) to help flag papers that may need more attention from area chairs (specifically, we flag papers with a large discrepancy between reviewer scores and self-rankings). More details on this will be given in a forthcoming blog post.
- Authors of accepted papers do not need to attend the conference in person. Their papers will still appear in the proceedings, but without a poster or a talk at the conference.
- Finally, we are planning to make an advanced reasoning LLM system available to ICML authors. Authors will be able to submit their paper drafts and receive feedback (up to a few days before the submission deadline). More details on this forthcoming. Stay tuned!
We hope that these efforts will continue to improve the quality of the conference and the ease of collaboration in our community.