arXiv Limits Researchers to Two Preprints per Month as Submissions Surge

To cope with the growing volume of manuscripts it receives, the preprint server arXiv has announced that, from 1 October, each submitting researcher will be limited to two papers per month.

The policy follows a rapid increase in arXiv submissions. Stein Sigurdsson, an astrophysicist at Penn State University in University Park and arXiv’s scientific director, said the rise is primarily linked to material generated by artificial-intelligence tools. In September, about 40,000 preprints were uploaded to arXiv, compared with roughly 30,000 per month from March to July.

Why arXiv is limiting submissions

Sigurdsson said the cap is the latest measure arXiv has introduced to prevent its roughly 300 moderators from being overwhelmed. Moderators currently check about 1,000 manuscripts per day in bulk.

The policy has also created a backlog. Sigurdsson said that although the move left “too many legitimate papers out there”, about 4,000 submissions are currently waiting to be published on the site.

Some researchers consider the restriction reasonable. Siddish Zadei, an epidemiologist at Columbia University in New York City, said the limit is not “inhibiting anyone’s productivity”, because preparing and posting two preprints a month already requires substantial effort.

Audit of 2.5 million biomedical papers reveals spike in false citations

Others argue that arXiv’s new policy should be stricter. Victor Galitsky, a theoretical physicist at the University of Maryland, College Park, said co-authors could bypass the limit by taking turns submitting their work.

Other preprint servers face similar pressure

SocArXiv, a social-science preprint repository, also limits authors to one submission per week. “There may be good reasons to upload three papers at once, such as dissertation chapters,” said Philip Cohen, a sociologist and demographer at the University of Maryland and director of SocArXiv. In most cases, however, he said that multiple submissions in quick succession are a “good hint for AI failure”.

Other venues are also dealing with a surge in submissions. Some major artificial-intelligence conferences and journals have adopted similar submission quotas.

Can submission limits reduce AI-generated research?

Some researchers say the problem of AI-generated or low-quality material cannot be solved simply by rejecting more articles. “If arXiv takes away these documents, they’ll just be moved somewhere else and become someone else’s problem,” said Reviewer 3, a platform operated in San Francisco, California, that uses AI to assist researchers with peer review.