AI Conferences Cap Research Paper Submissions as Applications Surge
AI conferences are flooded with applications.
Credit: MTStock Studio/Getty
Artificial intelligence conferences and journals are introducing limits on the number of papers that researchers can submit as they struggle with a surge in applications.
The 2027 International Conference on Learning Representations (ICLR), which will be held in San Francisco, California, has announced that each author can submit a maximum of 20 papers. Researchers who have never had a paper accepted at a major AI conference will be limited to one submission unless they have co-authors who meet the conference’s authorship requirements.
The policy comes as ICLR expects submissions to continue rising. The conference could receive more than 62,000 submissions, compared with 490 a decade ago.
“Historically, the majority of new researchers have focused their efforts on a single paper or project, and we want to encourage this practice rather than encouraging people to submit very large numbers of papers, which are typically generated by AI,” said Jacob Andreas, a computer scientist at the Massachusetts Institute of Technology in Cambridge and ICLR 2027 program chair.
AI conferences introduce submission quotas
The 2027 Annual Conference on Artificial Intelligence and Statistics will take place in Montreal, Canada, in May. Its new policy sets a limit of 12 submissions per author, falling to three submissions if the researcher has not had a paper accepted at a major AI conference. The change is intended to address the growing pressure on conference review systems.
The Association for Computational Linguistics has also set limits. Its conference, which will be held in Budapest later this month, will allow up to 20 submissions per person and five first-author papers per review cycle. The organization cited “unsustainable submission rates” when announcing the policy.
Researchers question whether the rules are fair
Not everyone supports the new restrictions. In a LinkedIn discussion about the ICLR changes, some researchers criticized what they saw as double standards.
“If a productive PhD student is required to submit two papers, why stop and allow 20 submissions from the lab, including previous ‘major conference’ papers?” said Irene Jenner, a computer scientist at The Ohio State University in Columbus, Ohio.
Abhinav Agarwal, a research engineer at Meta in Seattle, Washington, made similar criticisms, describing the restrictions on young researchers as “very disappointing.”
Kathleen Fraser, a computer scientist at the University of Ottawa in Ontario, Canada, said she “understands the logic” behind limiting submissions from researchers who cannot contribute to the review process. Under ICLR’s rules, only authors whose papers have been accepted at major AI conferences can review submissions.
Why research groups are affected
Gautam Kamath, a computer scientist at New York University who started the discussion on LinkedIn, thinks the 20-submission limit may be too high. However, he said the policy must accommodate large research groups, where a lab leader may be listed as an author on many papers.
Kamath is also co-editor-in-chief of the machine-learning journal Transactions on Machine Learning Research. In June, the journal introduced new submission quotas after submissions tripled over the previous year. The increase included a 13-fold rise in single-author submissions.
Source: www.nature.com