No single researcher can now push unlimited work through arXiv’s moderation queue. As of October 1, 2026, the ceiling is two papers per calendar month, per person.
An older limit stays in force: three active submissions may sit in flight at once, a rule arXiv has kept since 2024.
Whatever subject area a paper falls under, the cap trails the person who uploads it, and not only in computer science. Co-authors escape it entirely; only the submitter is charged with the slot.
A rejected manuscript burns a slot anyway, since it still costs moderators time. Pull a paper before it publishes, and the slot returns.
The figures driving the change are blunt. September 2016 brought 9,869 submissions; eight years on, that had reached 20,569. This September the total climbed to 40,363 and generated nearly 9,000 support tickets.
The AI category shows the steepest curve, its monthly volume multiplying more than sixfold across two years.
Thomas G. Dietterich, who chairs arXiv’s editorial advisory council, pointed to a small slice of authors drowning the queue in thin work. He named salami-sliced papers and dense machine-written text as the culprits crowding out serious submissions, and said sound papers now wait days or weeks.
He also wants the rule to discourage micro-results, the ones reporting a lone extra ablation or a minor architectural tweak.
arXiv calls the cap temporary, a stopgap until its moderation tooling catches up with the inflow.
