This page will be updated with questions (and answers) as they arise. Please check back regularly. If you do not find an answer to your question here, you are welcome to contact the IJCAI-25 program chair at pcchair@2025.ijcai.org, but please make sure that you have read IJCAI-25 CFP and the current FAQ first.
The reviewing process consists of two phases. In Phase 1, every paper will be reviewed by two reviewers and if a paper receives two reviews that are not sufficiently positive, it will be rejected without any opportunity to submit an author response. By submitting a paper, authors acknowledge that they are aware of the possibility of receiving a rejection notification after Phase 1.
Authors of papers rejected in Phase 1 will receive two reviews.
The IJCAI series of conferences started in 1969, and since then IJCAI has remained the premier conference bringing together the international AI community. In classic AI areas, with a long tradition at IJCAI, people submit their best papers to IJCAI. For areas that have grown a lot in the last few years (NLP, vision, machine learning), a rule of thumb to decide whether a paper is good enough for IJCAI, is to ask whether it is good enough for conferences such as ACL / ICCV / CVPR / NeurIPS / ICML / ICLR.
The purpose of the rebuttal is to ask questions about specific issues that 1) could directly influence your evaluation of the paper, and 2) do not require providing new results. Typical questions include requests to clarify or justify particular issues, or about important relationships to other works.
The purpose of the rebuttal phase is not to enter a discussion about the ways the authors could extend or revise their paper by, for instance, asking for extra experiments or theoretical results. You must judge the current submission, not the possible future versions of the paper.
All papers violating the rules specified in the CFP must be rejected. Most papers that obviously violated these requirements have been desk rejected before the reviewing phases. However, in case you find an obvious violation, please inform the workflow chairs (at workflow@2025.ijcai.org) immediately. You can also mention it in your review, though you must not take it into account in your judgment as decisions about violations are made by the PC-chair directly in order to ensure consistency and fairness.
Please refrain from using search engines such as Google to try to identify the authors. This is not good practice.
The rebuttal will be visible to all members of the program committee involved in the paper (PCs, SPCs, ACs, SACs, Associate PC Chairs, PC Chair).
There is also the option to enter confidential comments that will not be visible to the reviewers, but will be accessible to the meta-reviewers, senior meta reviewers, and (associate) PC chairs.
Please start by looking at our review form.
We highly recommend reading Mistakes reviewers make (by professor Niklas Elmqvist), see https://niklaselmqvist.medium.com/mistakes-reviewers-make-ce3a4c595aa2
It provides a very balanced view on reviewing that applies to AI and its subfields. (One little detail: when Elmqvist writes about user study (coming from an HCI perspective), you might want to think of experiment (in AI or ML).)
The program chair also highly recommends going through the slides that Hendrik Blockeel and Jesse Davis used for a tutorial on reviewing, and that was adapted for IJCAI-ECAI 2022, and is still highly relevant for IJCAI 2025.
https://dtai.cs.kuleuven.be/introduction-to-reviewing-for-ijcai.pdf
The assignment algorithm considers your bids, the subject areas, TPMS, as well as the reviewer workload, etc. It aims at optimizing the global assignment. If you did not bid, or entered positive bids for too few papers, or mainly on very popular papers, the algorithm may not have found the best papers for you. Still there are many other papers that need to be reviewed, and we rely on your expertise to provide an opinion.
Warning: If you see that a paper does not at all match the subject areas entered for that paper, please inform workflow@2025.ijcai.org and put it in your review, as this is considered bad practice.
If a paper is somewhat outside your core area of expertise, your evaluation of the work as an informed outsider can still be useful. Nevertheless, in exceptional cases where you believe you are unable to provide any well-informed review, please inform workflow@2025.ijcai.org as soon as possible. We will then try to assign it to someone else. Obviously, the longer you wait, the more difficult it is to reassign. (The same holds if you discover a conflict of interest, or if you cannot review a paper for any other reason.)
Please continue to review it. We don’t regard this situation as a COI. Reviewers for previous versions of the submission are in an ideal position to judge whether the current submission has been really improved. Note that submissions that have been rejected at any peer-reviewed conference within the past 12 months must declare the resubmission and attach the previous review as well as an explanation about how authors have improved the work. The resubmission file will be visible only after the rebuttal during the discussion period.
As mentioned in the call for papers, IJCAI 2025 welcomes papers on AI techniques for novel application domains. Keep in mind though that IJCAI is the place to publish research papers. Therefore, if the use of AI, or the AI technique used, is novel for this kind of application, it can be considered, provided that the application is formalized in such a manner that the formalization can be applied to solve other applications in an innovative manner as well. If the paper doesn’t completely satisfy the previous criterion but it clearly explains why the used technique works well, again it can be considered. In those cases, acceptance depends on how important the application is, how significant the improvement is, and to which extent it can be generalized to other problems. If the paper doesn’t satisfy the previous criteria but the AI technique used has a major impact in the application domain and in industry or society, it may be considered for publication at IJCAI 2025.
On the other hand, simply applying a standard AI or ML technique (like neural nets or decision trees) on standard datasets from an application domain clearly does not constitute a valid contribution for IJCAI 2025.
That is not a problem as long as the papers are presented in non-archival venues (that is, as long as there are not formally published proceedings with a publisher, in general with a DOI, ISBN, or ISSN). ArXiv submissions are also ok.
No, if the short paper is not longer than a two-column two-page paper. If the ‘short’ paper is longer than that, but no longer than half of the IJCAI paper, then the extension must be significant enough.