arXiv is rationing scientists to two papers a month, because AI has doubled the flood in two years

arXiv, the free preprint server where much of the world’s physics, maths and AI research appears before it is published anywhere else, will now accept only two submissions a month from each submitter. The repository announced the new limit on Thursday, saying AI tools have pushed it to a record 40,363 submissions in September.

That is double the 20,569 it received in September 2024, and four times the 9,869 of September 2016. The September flood alone generated almost 9,000 support tickets for arXiv’s staff and volunteer moderators.

Thin papers, ‘salami’ papers and dense AI prose

arXiv’s rules let authors use AI as long as they disclose it and the work meets its standards. The problem, it says, is that “many of the submissions that arXiv is now receiving do not satisfy these requirements.”

Its moderators are seeing more “thin papers of narrow scope”, more “salami” papers, where one piece of work is sliced into a set of smaller ones, and “a marked increase in dense, AI-written papers”. The sharpest growth is in its artificial intelligence category itself: submissions to cs.AI have risen more than sixfold in two years, while other categories have doubled.

Thomas Dietterich, who chairs arXiv’s Editorial Advisory Council, put the cost in plain terms:

A relatively small proportion of authors are submitting a large number of low-quality papers and consuming a disproportionate fraction of the moderators’ time. This is unfair to authors who continue to submit quality papers.

Thomas Dietterich, Chair of the arXiv Editorial Advisory Council

Good papers, he says, “can be delayed for days or weeks as a result.”

What changes for researchers

The new rules took effect on October 1 and apply across every subject area:

• Two a month: each submitter can make up to two submissions per calendar month, and no more than three can be active at any one time (the three-paper limit has been in place since 2024).
• Rejections count: the cap covers submissions, not accepted papers, because it is submissions that use up moderators’ time.
• Only the submitter is counted: co-authors aren’t affected when someone else submits a joint paper, so teams will need to coordinate who submits what.
• Some exceptions: papers still on hold from a previous month don’t count against the new month’s allowance, and a paper deleted before it is announced doesn’t count at all.

arXiv calls the limit a “stopgap” while it works out “the new best practice for authors employing more and more advanced AI tools” and improves its moderation tools. Researchers with a backlog, or a conference deadline, are told to plan their submissions carefully.

Why it matters

arXiv is where new AI research, including the papers behind stories like the Pain Axis study, appears first, often months before peer review. When AI-generated volume forces one of science’s most open doors to start rationing access, it shows the same tools mathematicians are already wrestling with are now straining the system that shares research with everyone else.

Source: arXiv blog.