ICLR 2027 Receives 50,000 Abstracts Ahead of Deadline

The ICLR 2027 conference has already recorded approximately 50,000 abstracts, even though the submission deadline has not yet been reached, far surpassing the volume of the 2026 edition. The difficulties related to the use of AI in submissions and evaluations, observed last year, have undermined confidence in the peer review process and could intensify with such an influx.

The Failures of 2026 Fuel Fears of a New Wave of Complaints

During the 2026 edition, ICLR faced low-quality submissions generated by AI, as well as evaluations produced using these tools. Some submitted papers included fabricated citations, and reviewers resorted to AI to cope with the influx of texts. These practices contributed to a loss of trust in the peer review process. With an even higher number of papers expected this year, it is likely that complaints regarding quality and evaluation will multiply.

A Colossal Volume for 2027, Exceeding the 19,500 Submissions of 2026

ICLR 2027 has already received around 50,000 abstracts, while the submission period is still ongoing, compared to about 19,500 valid submissions for the previous edition. Some of these abstracts likely come from authors awaiting the results of NeurIPS and may withdraw their submission to ICLR if their work is accepted elsewhere. Therefore, the final number of submissions will be lower than the current volume, but it should remain significantly higher than last year's.

Multiple Drivers, Including AI-Enabled Writing Speed

Several factors explain this increase in submissions: the general interest in AI, the rise in corporate research budgets, where compensation can sometimes depend on the number of publications, and especially the ability of AI to accelerate article writing. An analysis conducted on NeurIPS showed that authors widely utilized AI to draft their submissions.