Preprint server
Appears in 6 stories · Last updated Oct 19, 2026
Has imposed new submission rate limits
In October, arXiv capped each researcher at two submissions per month after AI tools drove September submissions to a record 40,363. The policy, effective October 1, also limits active submissions to three. Nearly 300 volunteer moderators check about 1,000 papers a day.
Updated Oct 19
As of September 2026: Hosted the proof paper on September 14, 2026
A paper posted to arXiv on September 14 claims to settle the k-server conjecture, a problem in theoretical computer science open since 1988. Authors Christian Coester, Elias Koutsoupias, and Marek Zbysiński say they have proven that the Work Function Algorithm achieves the optimal competitive ratio of k on every metric space.
Updated Sep 15
As of September 2026: Where the paper was posted
Two high school students from Oak Park High School in California, working with a UCLA postdoc, posted a paper that appears to finish a problem Fields Medalist June Huh worked on without closing. It describes, for any Lorentzian polynomial of any degree in any number of variables, exactly which coefficient ratios are universally bounded — and the tightest possible bound in the three-variable case.
Updated Sep 14
As of September 2026: Hosts the RLT report and related recurrent-depth research
Yifan Zhang, a Princeton researcher, posted a technical report September 12 describing the Recurrent Looped Transformer (RLT). It pairs a causal encoder with a recurrent decoder that carries its hidden state and sliding-window attention cache across every token.
Updated Sep 13
As of September 2026: Hosted the discovery paper
Astronomers have caught a wandering black hole feeding from its own wake. The 35,000-solar-mass object sits outside the star-forming disk of the dwarf galaxy UGCA 320, about 20 million light-years from Earth, pulling gas from the dense trail its gravity leaves behind.
As of September 2026: Hosted the outage study
A preprint posted September 2 asked whether an off-the-shelf language model, trained on no utility data at all, can predict which storms cause the worst power outages. Four zero-shot LLMs were benchmarked against two supervised machine learning models on six years of central Texas outage records.
Updated Sep 8
No stories match your search
Try a different keyword
How would you like to describe your experience with the app today?
