ICLR and arXiv Cap Paper Submissions Amid AI Surge

Major academic venues in machine learning are enacting strict restrictions to curb an overwhelming surge in research submissions. The International Conference on Learning Representations (ICLR), which handled roughly 500 papers a decade ago, is now projecting over 40,000 submissions for its 2027 event. Open-access research-sharing repository arXiv experienced a similar boom, receiving over 40,000 submissions in September alone, which is nearly double the number it received two years ago.

To manage the critical shortage of qualified reviewers caused by growing field interest and AI tools that make paper writing faster, platforms are forcing structural changes. ArXiv is now limiting authors to two papers per month. Meanwhile, both ICLR and the Association for Computational Linguistics’ Rolling Review (ARR) system are introducing a cap of 20 papers per author for each submission period. To resolve severe bottlenecks, ARR will also use a lottery to decide which papers without qualified reviewers get reviewed.

Other top organizations are targeting the peer-review process directly. NeurIPS is cracking down on poor-quality reviews by requiring researchers to finish their assigned reviewer duties before they can see feedback on their own submitted work. Furthermore, AI-written papers remain allowed across major venues because AI detectors are not reliable enough and researchers have not agreed on how much AI use is acceptable.

For practitioners and academics, these policy shifts alter how research is published and evaluated. While the goal is to cut down on low-quality submissions and make reviewer workloads manageable, some researchers worry these strict submission limits and randomized reviewing elements could make things harder for younger academics trying to build a career.