ICML 2025 Peer Review FAQ
This page addresses some frequently asked questions we have received regarding the peer review process for ICML 2025. Due to the high volume of emails we receive, we may not be able to respond to inquiries, especially to those that are already addressed on this page.
General information about the author response and author-reviewer discussions
- Number of reviews: Over 99% of submissions have at least three reviews at the start of the author response period. However, despite our best efforts, a small number of papers still do not have three reviews. Emergency reviews are being obtained for these papers, and will be added in the upcoming days. When needed we will step in for these papers ourselves, ensuring that they receive proper attention.
- Some submissions have more than 3-4 reviews. Authors can ask for guidance on which reviews to focus on in their author response by leaving a confidential comment for ACs.
- Review scores: As a general note, what eventually matters is the review text and not the final score. The ACs will integrate all information from the reviews and discussion into their final decision. Thus, the average score should not be taken as a direct indication of the final decision on the paper.
- Discussion strategy: Past experience suggests that effective responses focus on factual errors in the reviews and on responding to specific questions posed by the Reviewers. Your response is optional and should be reserved for cases when a response is called for. The response you submit by March 31st should contain all your arguments regarding the reviews. Additional discussion with reviewers should be reserved only for discussing these points, rather than raising new arguments.
- Cannot Update Submission: Similar to previous years, the original submission (PDF and supplemental material) cannot be revised in OpenReview during the discussion period.
- Links and Anonymity: Your responses to reviewers should not contain or link to any identifying information that may violate the double-blind reviewing policy. While links are allowed, reviewers are not required to follow them, and links may only be used for figures (including tables), proofs, or code (no additional text). All links must be anonymous to preserve double-blind review, both in the URL and the destination.
- Making the reviews and author discussion public: New this year, for accepted papers (and for rejected papers that opt-in) reviews and responses you post to the reviewers will be made public at the end of the review process. Remember this when posting your messages, and don’t write anything you would not like to be made public.
- Professionalism: Throughout this process, please be respectful and professional in your dialogue, and abide by the ICML code of conduct: https://icml.cc/public/CodeOfConduct
Email communications
Q: From whom should I expect emails about the ICML peer reviewing process?
A: Emails via OpenReview will come from noreply@openreview.net and icml2025-notifications@openreview.net. Please make sure those emails are able to make it past any email/spam filters.
(Unfortunately these email addresses were accidentally omitted in a communication to authors.)
Q: I have a question not answered on this page: [...]?
A: The (general) ICML FAQ page answers a number of questions about icml.cc accounts, conference logistics, payment, registration, travel documents, etc. If you still cannot find the answer to your question, please use the following form to direct your question to the most appropriate organizer: https://icml.cc/Help/Contact. We will try our best to get back to you or update this page with an answer as soon as possible.
Post-submission updates
Q: I accidentally { failed to upload a PDF, uploaded the wrong PDF, used the "accepted" option when using icml2025.sty, forgot to include the appendix, left out one or more co-authors, introduced PDF font or encoding errors, included my GitHub username/password in the PDF or supplementary material, ... }. Can I upload a (new) submission after the deadline?
A: No, it is logistically infeasible to accommodate these requests; it is also unfair to other authors and would set a bad precedent. Please also see below regarding requests for changes to the author list.
Q: ... What if I forgot to include the (required) Impact Statement? Will my submission be desk-rejected?
A: See answer below.
Changes to the author list
Q: I forgot to add a co-author to my submission. Can I add the co-author now, even though it is after the abstract submission deadline?
A: No. At the time of submission, the submitting author explicitly checked a box affirming that "[they] have double-checked the author list and understand that additions and removals will not be allowed after the abstract submission deadline." If the authorship of the submission is incorrect or unacceptable, then the submission should be withdrawn.
This also applies if a co-author was omitted due to not activating their OpenReview account in time (as described in the Call for Papers).
Conflict of Interest (COI)
Q: How do I update Conflict of Interest (COI) information in OpenReview?
A: Sign into your OpenReview account, go to your profile page, and switch to edit mode by clicking "Edit Profile" near the top. You should arrive at a page titled "Edit Profile", and you can switch between the various sections by clicking the numbers near the top of the page (corresponding to Names, Personal Info, Emails, Personal Links, History, Relations, and Expertise). Of particular relevance are Emails, History, and Relations. See https://icml.cc/Conferences/2025/ConflictOfInterestDefinitions for what is considered a COI for ICML 2025. Please also make sure the "Emails" part of your profile lists your institutional email addresses (e.g., from your university, lab, or workplace).
(We sent a reminder to all authors to update this information; don't worry about the reminder if everything is already up to date!)
Q: I have a lot of co-authors, and it will take too much effort to enter them all as "relations" in my OpenReview profile. Is there anything I can do to reduce this effort?
A: Co-authors of your publications that appear in your OpenReview profile do not need to be explicitly entered as "Relations" in your profile. (OpenReview allows you to quickly import publications listed in your DBLP page; this can be accessed from the "Personal Links" section when you edit your profile.)
Submitting a paper
Q: I missed the abstract (or full paper) submission deadline because of <insert very good reason>, can I get an extension?
A: Sorry, these deadlines are strict with no exception.
Q: I have a paper under submission to another conference. Is it okay under the Dual Submission Policy to also submit it to ICML 2025?
A: In order to comply with the ICML Dual Submission Policy, you must ensure that a paper of yours accepted or under review at another conference is not under submission to ICML 2025 by the full paper submission deadline. Note that the ICML 2025 full paper submission deadline is after the notification dates for both AISTATS 2025 and ICLR 2025, so you should be able to remove any potential dual submission from OpenReview in time.
Q: Is it possible to only submit an abstract, or to only submit a poster without a full paper?
A: No. We are only taking submissions for full (research) papers and position papers.
Q: Is it possible to submit a full paper without an abstract submission?
A: No. An abstract must first be submitted by the Abstract submission deadline on OpenReview. After submitting the abstract, it will be possible to modify the submission to also upload the full paper PDF, which must be uploaded by the Full paper submission deadline. See the Call For Papers for further details (e.g., what intervening modifications are allowed).
Q: Is it possible to present a paper virtually or in hybrid-mode?
A: The conference is planned to be an in-person event, with no support for virtual or hybrid presentation. Barring exceptional circumstances, we require at least one author of each accepted paper to register for and attend the conference to present the paper.
Q: Which license should I choose?
A: Please see the OpenReview legal terms. (The arXiv license information page has related advice.)
Q: I’m confused, what is actually the file size limit for the submission PDF?
A: As mentioned in the author instructions, we use the standard OpenReview limit of 50MB for the submission PDF. For the camera ready PDF, we will require a file size limit of 20MB. Earlier versions of the example paper mentioned a limit of 10MB, but this is outdated information.
Q: Can I modify the ICML LaTeX template (e.g., by commenting-out <insert LaTeX code here, like \printAffiliationsAndNotice>) so that I have more space for something else?
A: Modifying the template to gain an unfair "space" advantage relative to other authors is not allowed. However, we decided not to desk-reject papers for commenting-out \printAffiliationsAndNotice due to confusion in the LaTeX template (Updated: March 18, 2025).
Q: Is (or was) the Impact Statement required at the time of submission?
A: Yes -- see the Call For Papers. However, we decided not to desk-reject papers for failing to include an impact statement. Please note that reviewers are asked to comment on ethical concerns in their reviews, and the lack of an impact statement may be taken into account by the reviewers. (Updated: February 6, 2025.)
Double-blind reviewing
Q: I want to include a link to my code/data in my submission, but it includes information (e.g., a github username) that could reveal my identity. Does this violate the double-blind review policy?
A: Yes, this would violate the double-blind review policy. If you include a link to a code or data repository that includes any identifying information about you, your submission may be desk-rejected. Please anonymize any code/data/links that you include in your submission.
Q: Is it okay to search the internet for a submission that I am assigned to review? Is it okay that I have already seen a submission that I am assigned to review (and hence already know who the authors are)?
A: Reviewers should NOT search the internet (or elsewhere) for submissions they are assigned to review, as this could violate the double-blind review policy. Naturally, this is not something we can explicitly enforce, and reviewers may have legitimately already seen de-anonymized versions of submissions they are assigned (e.g., if a literature search turns up an arxiv preprint). But the goal of the double-blind review policy is to try to reduce biases that might arise from having knowledge of a submission's authors. Also reviewers should wait to have read the paper at least once before doing a literature search on the topic of the paper.
Reciprocal reviewing requirements
Q: How do I sign up as a reviewer to fulfill the Reciprocal Reviewing Requirement?
A: There are two Reciprocal Reviewing Requirements: a "per-submission" requirement, and a "per-reviewer" requirement. The abstract submission form (on OpenReview, starting January 9) will allow submitters to designate an author to fulfill the "per-submission" requirement, or to indicate that the submission is exempt from the requirement. Any author with 4 or more submissions should fill out this Google Form to provide information we ask of all reviewers, or to indicate that they are exempt from the requirement.
After the abstract submission deadline, authors serving as reviewers to fulfill Reciprocal Reviewing Requirements may be contacted through OpenReview via <noreply@openreview.net> in that service capacity.
Q: Is the {first, second, ..., last, junior, senior, advising, sponsoring} author automatically exempt from the Reciprocal Reviewing Requirementement? How about authors who are in industry or who have left academia?
A: All authors are treated the same in terms of the Reciprocal Reviewing Requirement.
Being an area chair or reviewer
Q: I would like to volunteer to be an area chair or reviewer for ICML 2025. How can I do this?
A: Please fill out the following Google Form: https://forms.gle/n4SZYcxKopQoHnp27. (It will close December 27, 2024.) You will need to provide your OpenReview profile as part of this form submission.
This form may also be used to register yourself as a reviewer to fulfill the Reciprocal Reviewing Requirement for submissions you have co-authored. (After December 27, we will use a different process for this, which will be available by the time the submission site is open on January 9.)
A: We are no longer accepting (self-)nominations for area chairs or reviewers.
Q: What are the qualifications to be a reviewer for ICML?
A: Reviewers must have research experience equivalent to a second-year graduate student in machine learning or a related field. They must have been a primary author* on at least two peer-reviewed conference or journal papers published in a related venue (e.g., ICML, NeurIPS, ICLR, UAI, AISTATS, COLT, ALT, JMLR, TMLR, CVPR, ICCV, ACL, NAACL, EMNLP, SIMODS – note that this is not meant to be an exhaustive list). We strongly encourage each first-time reviewer to identify a ‘mentor’ (such as a research advisor or manager) who has both the necessary qualifications for and prior experience with reviewing, and who has agreed to oversee and assist the reviewer in their reviewing tasks.
The research experience ensures the reviewer is to be able to competently evaluate a submission’s methodology, interpret findings and results, and to evaluate contributions in the context of prior works. Prior authorship ensures that the reviewer understands the peer review process (at least from the side of the authors) and the standards and conventions of composing reviews and corresponding with authors.
*We leave it to your own discretion to interpret what is meant by "primary author", as this may vary between sub-areas of machine learning.
Q: What are the qualifications to be an area chair for ICML?
A: ACs (a.k.a. meta-reviewers) must have seniority at least at the level of a junior faculty member (e.g., assistant professor) in an academic institution or industry equivalent. They must have substantial prior experience with reviewing for peer-reviewed conferences or journals. They must also have expertise and/or broad knowledge in multiple major sub-areas of machine learning.
The seniority and reviewing experience ensure that the AC is able to oversee and ensure the quality of work of several reviewers and provide feedback/guidance to reviewers where necessary. The expertise and broad knowledge ensure that the AC is able to judge the contributions and importance of the submissions relative to prior works and ultimately make recommendations about acceptance/rejection.
Q: What is the reviewing load for reviewers? Can I get a reduced reviewing load?
A: We are aiming for a load of around five submissions per reviewer. Unfortunately we are not allowing for reduced load reviewing this year.
Q: As an AC, how do I communicate with authors?
A: You can use the “Author AC Confidential Comments” button.
Generative AI policy
Q: Why does ICML prohibit the use of Generative AI tools (including LLMs) in reviewing?
A: Privileged information—such as the submission themselves, reviews of the submissions, discussions about the submissions, etc.—may only be used for the purpose of reviewing. Therefore, we cannot allow this information to be submitted to external services such as Generative AI tools.
The reviewing process is meant to obtain reviews from peer researchers who have read, understood, and evaluated the correctness and merits of the submissions. The credibility of ICML is damaged if reviews are automatically produced using Generative AI tools such as LLMs (whether they are accessed via an external service or on a local machine).