THE CRUNCH
The International Conference on Learning Representations, or ICLR, has already pulled in about 50,000 abstracts for its 2027 edition, with the deadline still a week away. This figure dwarfs the 19,500 valid submissions received for ICLR 2026. The Decoder reports that the surge is likely driven by the broader AI hype and corporate research spending, where publication records can influence pay. Additionally, authors of
The Decoder notes that AI tools make it much faster to write papers, and a NeurIPS analysis found authors used AI heavily to draft submissions. This has already caused problems for ICLR 2026, which struggled with low-quality AI-generated papers and fabricated citations. Reviewers also turned to AI to cope with the volume, eroding trust in the peer review process. With even more papers flooding in this year, complaints about the quality of the conference are expected to grow louder.
The Decoder suggests that some abstracts may be hedging bets, with authors waiting on NeurIPS results before withdrawing if accepted there. This means the final count will likely be lower, but still far above last year's total. The flood of submissions highlights the intense competition in the AI research community and the pressure to publish.