arXiv Implements Submission Limits Starting October 1: Each Author Limited to Two Papers per Month

Download the app

+1 Flood of AI "junk papers" prompts arXiv to impose submission limits Wang Fang / China Science Daily 2026-10-22 07:39 Science Pulse > Listen to full text Font size

According to a report in Nature, to cope with the challenge of a surge in submissions, the preprint platform arXiv has officially implemented submission limits starting October 1 — each author may submit at most two papers per month.

Steinn Sigurðsson, arXiv's scientific director and a professor at Pennsylvania State University in the United States, said the surge in submissions mainly stems from the large number of "junk papers" produced by artificial intelligence (AI) tools. In September, authors uploaded 40,000 preprint papers to the arXiv platform, compared with about 30,000 per month from March to July.

Over the past year, arXiv has introduced multiple measures to control submission chaos. Sigurðsson said that last October the platform tightened submissions in the computer science field, because that section had become a disaster area for fraud. In January this year, the platform updated its author eligibility requirements: authors who want to upload preprint papers need to have a prior publication record or be recommended by a senior arXiv user, and provide an institutional email address. In May, the platform banned a group of authors whose papers had reference hallucination problems.

Sigurðsson said arXiv launched this new measure to protect the platform's nearly 300 volunteer reviewers from being overwhelmed by the massive volume of manuscripts. The review team now handles about 1,000 manuscripts per day on average. But he acknowledged that the submission limits have also slowed the release of many legitimate, high-quality papers. Currently, nearly 4,000 manuscripts are still queued on the platform awaiting public release.

Regarding this restrictive measure, some scholars consider it entirely reasonable. Siddhesh Zādéy of Columbia University in the United States said the move "will not suppress scientific output," because producing two preprint papers per month itself requires considerable work. At the same time, some scholars believe the new arXiv rules are still not strict enough. Victor Galitski of the University of Maryland, College Park, warned that co-authors could take turns submitting to circumvent the limits.

Another preprint platform, SocArXiv, already stipulates that each author may submit at most one paper per week. Philip Cohen, head of SocArXiv and a professor at the University of Maryland in the United States, said: "There are indeed times when uploading three papers at once is reasonable. But in most cases, submitting multiple papers in a short period is often a signal of AI 'junk papers.'"

Some views hold that simply rejecting papers cannot solve the AI "junk paper" problem. "Even if arXiv rejects these manuscripts, they will flow to other platforms and become someone else's problem," said Natalie Khalil, founder of the US AI peer review platform Reviewer3. Preprint platforms should find ways to strengthen the processing capacity of verification and review, for example by using tools to check the reproducibility of paper code, validate references, and check manuscript metadata.

arXiv officials likewise believe AI technology can be used to counter AI-generated "junk papers." Thomas Dietterich, head of arXiv's computer science section and a professor at Oregon State University in the United States, said: "We are testing some review technologies based on large language models, but they are still far from actual deployment and application." Responsible editor: Huan Yanhong Image editor: Chen Feiyan The Paper news tip: 021-962866 The Paper, unauthorized reproduction prohibited +1 Favorite I want to report #ScientificResearch #Papers See more