arXiv limits researchers to two submissions a month after an AI-fuelled surge

[Tiny Kernel]

arXiv, the preprint server that much of physics, maths and computer science relies on, has introduced a cap on submissions. Each submitter can now send two papers per calendar month, with a maximum of three active submissions at any one time. Times Higher Education reports that the policy is dividing opinion, and it is easy to see why. For decades arXiv has been open to almost anyone with a relevant paper

The numbers explain the decision. arXiv received 40,363 papers last month, compared with 9,869 in September 2016, and submissions have roughly doubled in just two years. All of those papers have to be checked by a small staff and around 170 volunteer moderators, who make sure they are in the right category and meet basic standards. A lot of people link the recent jump to AI tools making it easier to churn out papers. Volunteers can only do so much

Editor in chief Thomas Dietterich said some authors are submitting dozens of papers, which takes moderation time away from other researchers and causes delays. The cap is meant to share volunteer attention more fairly and protect the collection from inappropriate submissions. Rutgers mathematician Alex Kontorovich backed it, telling researchers to pick two, their very best two. Astronomer Catherine Heymans also spoke up for quality over quantity

Not everyone agrees. Harvard's Scott Kominers argued that arXiv should economise on moderation, not on scholarship, saying the policy confuses the two. Software engineer Dare Obasanjo summed up a common feeling, calling it another open submission process killed by LLMs. There are also worries about large collaborations and prolific researchers whose output is legitimate

We have linked plenty of arXiv papers on this forum, so this affects the kind of research we see. I can understand the problem, but a blunt cap feels like it punishes everyone for the actions of a few. Is two a month fair? Or should arXiv be using better tools to filter low quality papers instead?. Researchers outside the field also rely on it more than they might realise

arXiv's new preprint submission cap divides opinionwww.timeshighereducation.com

[TheLegendKyle21]

There is a slightly funny contradiction here: arXiv exists partly to make research move faster, and now it has to slow researchers down because the tools for producing research-shaped documents have become too efficient

Still, I would not assume every increase in submissions is caused by AI. Computer science has been expanding rapidly for years, and more researchers naturally means more papers. AI may be adding noise, but it is not the only possible explanation.

The useful statistic would be the change in submission volume, moderation time, and rejection or withdrawal rates. Without those numbers, it is difficult to know whether the new limit is solving a major problem or just buying the moderators some breathing room.

Either way, two submissions is a blunt instrument. It is simple to enforce, though, and simple rules have a certain appeal when the alternative is building an elaborate automated gatekeeper.