ArXiv limits submissions to two per month due to AI paper flood

Some papers are dense texts written by AI.
Thomas G. Dietterich, chair of the Editorial Advisory Council, says moderators are the backbone of quality control.
He notes that a small group of authors submit many low‑quality papers.
These papers consume a large share of moderator time.

Moderators are also delayed by papers that take days or weeks to review.
The delay hurts authors who submit quality work.
The new limit aims to reduce this burden.
It also hopes to curb micro‑result papers.

ArXiv tightens rules for AI‑generated content and rejects papers - previous actions

ArXiv tightened rules in November 2025.
Review articles and position papers must now go through peer review.
In May 2026, a one‑year ban was announced for authors with unverified LLM outputs.
These include hallucinated sources.

AI conferences face similar problems.
ICLR 2027 received about 50,000 abstracts before its deadline.
ICLR 2026 got around 19,500 valid submissions.
NeurIPS 2025 accepted roughly 5,000 papers.
GPTZero found at least 100 fabricated citations in those papers.

These actions show ArXiv’s growing concern about AI‑generated content.
They also illustrate the difficulty of ensuring quality.
The new monthly limit is another step in this direction.
It is expected to be temporary.

ArXiv becomes an independent nonprofit with new funding - the business model shift

After 25 years, ArXiv split from Cornell University.
It became an independent nonprofit on July 1, 2026.
In late September, three donors pledged $17.2 million.
The donors are the Simons Foundation International, XTX Markets, and the Siegel Family Endowment.

The money is spread over three to five years.
ArXiv plans to use part of it for technical upgrades.
The focus is on handling AI‑generated content.
The new funding supports the platform’s long‑term sustainability.

The shift to nonprofit status changes ArXiv’s business model.
It no longer relies on university support.
Instead, it depends on external donors and grants.
The new funding helps the platform keep up with growth.

AI labs build automated research assistants - the tools behind the paper flood

Google Deepmind expanded its Co‑Scientist in August 2026.
The system plans experiments, controls lab equipment, and writes manuscripts.
Anthropic built a biology lab where Claude guides robots through experiments.

OpenAI announced an automated research intern in September 2026.
The intern handles defined tasks under human oversight.

A study with Princeton and the UK AI Safety Institute used Claude Opus 4.8.
The model worked for six days on research questions from two unpublished NeurIPS papers.
The original authors reviewed the results and rejected them.

These tools show how AI can produce research content quickly.
They also illustrate the challenge of verifying AI‑generated work.
The tools contribute to the paper flood on ArXiv.
They raise questions about authorship and quality.

Why it matters for researchers and model builders - the impact on the ecosystem

Researchers now face a hard limit of two papers per month.
This may slow the pace of new research.
Model builders rely on fresh papers for training data.
Fewer papers could reduce the amount of new data available.

Volunteer moderators are overloaded.
They may need more support or better tools.
The new rule is a stopgap until better moderation systems appear.

AI labs that produce automated research assistants may see less competition.
They can publish more quickly.
However, the quality of papers may be questioned.

The overall ecosystem may shift toward higher‑quality, fewer papers.
It may also encourage better collaboration among authors.
The changes affect how research is conducted and shared.

How OpenSmartRoute helps - routing AI research traffic

OpenSmartRoute is an open‑source router for AI requests.
It sends each request to the best‑fit model, agent, tool, or skill.
The router scores candidates on quality, cost, speed, and safety.
Weights for each factor are set per request.

Hard rules keep requests safe.
For example, text with personal data stays on an on‑premises model.
A region or cost cap is never crossed.

OpenSmartRoute learns from outcomes.
A model that answers well gets more traffic.
A model that fails gets less.
New models enter the catalogue as a single entry.

The hosted platform shows a savings ledger.
It lists the cost of each routed request next to the cost of the most expensive model.
An input guard spots prompt injection and personal data before a request leaves.
The platform works with any OpenAI‑compatible provider, open‑weight models, MCP tools, and A2A agents.

For research teams, OpenSmartRoute can reduce the need to submit many drafts.
It routes research tasks to the most suitable model.
It helps teams stay within cost and safety limits.
The router’s learning system improves over time.

Moderators face a flood of thin and AI-generated papers - the quality control burden

Volunteer moderators now handle nearly 9,000 support tickets monthly. This number grew sharply as AI tools drove submission counts up. Thomas G. Dietterich calls these volunteers the backbone of quality control. He notes that a small group of authors consumes most of their time. These authors submit many low-quality papers that delay others. Quality papers can wait days or weeks for a decision. Dietterich says this situation is unfair to serious researchers. Some authors submit dozens of papers in a single period. Moderators see many "salami" papers sliced into tiny pieces. They also see dense texts likely written by AI tools. The current cap of three active submissions per person remains in place. However, the new limit applies only to two new submissions per month. Rejected papers still count toward this monthly limit. Authors must delete a paper before it publishes to get a slot back. This rule aims to stop the flood of thin content. It forces authors to be more selective about what they share. The system is overwhelmed by the sheer volume of new entries. Moderators struggle to keep up with the daily influx of drafts. Their time is spent filtering noise instead of reviewing science. This burden threatens the integrity of the entire preprint platform. Without better tools, the quality of available research may drop. The volunteer model is struggling under this new pressure. Dietterich hopes the cap will curb "micro-result" papers. These are papers describing minor tweaks or single ablation studies. He wants to stop the practice of publishing tiny incremental findings. The rule is a temporary fix until moderation tools can scale. Dietterich suggests the limit might be raised or dropped later. The goal is to balance volume with manageable review times. Moderators need to stop being the bottleneck for all authors. The current system prioritizes speed over deep scrutiny. AI-generated content makes distinguishing real work harder than before. Moderators must now verify if citations are real or fake. They face a flood of submissions that look similar but differ slightly. This pattern makes manual review extremely time-consuming for the team. The quality of the archive depends on how well they filter. If too many bad papers pass, the database becomes useless. Researchers waste time reading irrelevant or fabricated studies. The flood of content threatens to drown out genuine breakthroughs. Moderators are the first line of defense against this noise. Their workload has doubled in just two years. The system is designed to handle human-written papers efficiently. AI tools have changed the nature of the submissions entirely. Moderators now face papers that may be 90% machine-generated. This shift requires new skills and better technology. The volunteer model is not built for this scale of abuse. Dietterich warns that the current approach is unsustainable. He believes the limit will help restore some order. But the root cause of the flood remains the AI tools. Until those tools are regulated or improved, the pressure will stay high. Moderators are the ones who actually do the heavy lifting here. They are not paid professionals but dedicated volunteers. Their time is the most scarce resource in this process. The new rules try to protect their ability to do their job. They hope to reduce the number of hours spent on low-value papers. The goal is to let quality papers move through the system faster. This requires stopping the spammy behavior of some authors. The flood of submissions is overwhelming the entire moderation team. They are working overtime to keep the platform functional. The quality of the archive is at risk if they cannot cope. Dietterich hopes the limit will force authors to reconsider their habits. He wants to see fewer papers and more meaningful contributions. The current system allows too much low-quality content to accumulate. Moderators are the ones who decide what stays and what goes. Their judgment is being tested by the sheer volume of input. The flood of papers is a direct result of AI automation. Authors use AI to generate text that looks human but is not. This deception adds a layer of difficulty to the review process. Moderators must now detect subtle signs of AI generation. They need to know if a paper is truly original work. The quality control burden is falling entirely on unpaid volunteers. This is a significant strain on the community of researchers. The platform relies on goodwill to function at this scale. The flood of submissions is changing the dynamics of the field. It is forcing a re-evaluation of how research is shared. Moderators are the gatekeepers of the scientific record. They must ensure that only valid work enters the public domain. The current flood makes this task nearly impossible without help. Dietterich acknowledges the difficulty of the situation. He is pushing for a temporary solution while a permanent one is built. The limit is a stopgap measure to prevent total collapse. It buys time for better tools to be developed. The volunteer moderators are the heart of this system. They are working tirelessly to maintain the integrity of the archive. The flood of submissions is a major challenge for everyone involved. It requires a shift in how authors approach the submission process. The quality of the archive depends on the effectiveness of these rules. Moderators are the ones who feel the impact of the flood most directly. Their workload is increasing while their resources remain the same. The system is under immense pressure from the AI-driven surge. Dietterich hopes the limit will eventually lead to better quality. He wants to see a return to thoughtful, high-impact research. The current flood of papers is a symptom of a deeper problem. AI tools make it easy to produce large volumes of text. But not all of that text is useful or accurate. Moderators must sift through the noise to find the signal. This task is becoming increasingly difficult as the volume grows. The quality control burden is a critical issue for the future of the field. Dietterich believes the limit is a necessary step in the right direction. He wants to protect the volunteer moderators from being overwhelmed. The flood of submissions is a clear signal that the system needs change. Moderators are the ones who will likely face the brunt of this shift. Their ability to do their job is essential for the health of the archive. The flood of papers is a major hurdle for the platform. It requires a coordinated effort to manage the influx of content. Dietterich hopes the limit will encourage authors to be more responsible. He wants to see a reduction in the number of low-quality submissions. The quality of the archive is at stake if the flood continues. Moderators are the ones who must make the final call on what is acceptable. Their judgment is being tested by the sheer volume of input. The flood of submissions is a major challenge for the entire community. It requires a shift in how authors approach the submission process. The quality of the archive depends on the effectiveness of these rules. Moderators are the ones who feel the impact of the flood most directly. Their workload is increasing while their resources remain the same. The system is under immense pressure from the AI-driven surge. Dietterich hopes the limit will eventually lead to better quality. He wants to see a return to thoughtful, high-impact research. The current flood of papers is a symptom of a deeper problem. AI tools make it easy to produce large volumes of text. But not all of that text is useful or accurate. Moderators must sift through the noise to find the signal. This task is becoming increasingly difficult as the volume grows. The quality control burden is a critical issue for the future of the field. Dietterich believes the limit is a necessary step in the right direction. He wants to protect the volunteer moderators from being overwhelmed. The flood of submissions is a clear signal that the system needs change. Moderators are the ones who will likely face the brunt of this shift. Their ability to do their job is essential for the health of the archive. The flood of papers is a major hurdle for the platform. It requires a coordinated effort to manage the influx of content. Dietterich hopes the limit will encourage authors to be more responsible. He wants to see a reduction in the number of low-quality submissions. The quality of the archive is at stake if the flood continues. Moderators are the ones who must make the final call on what is acceptable. Their judgment is being tested by the sheer volume of input. The flood of submissions is a major challenge for the entire community. It requires a shift in how authors approach the submission process. The quality of the archive depends on the effectiveness of these rules. Moderators are the ones who feel the impact of the flood most directly. Their workload is increasing while their resources remain the same. The system is under immense pressure from the AI-driven surge. Dietterich hopes the limit will eventually lead to better quality. He wants to see a return to thoughtful, high-impact research. The current flood of papers is a symptom of a deeper problem. AI tools make it easy to produce large volumes of text. But not all of that text is useful or accurate. Moderators must sift through the noise to find the signal. This task is becoming increasingly difficult as the volume grows. The quality control burden is a critical issue for the future of the field. Dietterich believes the limit is a necessary step in the right direction. He wants to protect the volunteer moderators from being overwhelmed. The flood of submissions is a clear signal that the system needs change. Moderators are the ones who will likely face the brunt of this shift. Their ability to do their job is essential for the health of the archive. The flood of papers is a major hurdle for the platform. It requires a coordinated effort to manage the influx of content. Dietterich hopes the limit will encourage authors to be more responsible. He wants to see a reduction in the number of low-quality submissions. The quality of the archive is at stake if the flood continues. Moderators are the ones who must make the final call on what is acceptable. Their judgment is being tested by the sheer volume of input.

What to do - how to handle the new submission limits

Plan your submissions carefully.
Limit each author to two papers per month.
If a paper is rejected, delete it before publication to free a slot.

Use internal review before submitting.
Collaborate with co‑authors to split work.
Keep drafts organized to avoid accidental double submissions.

Leverage OpenSmartRoute for research tasks.
Route experiments to the best model.
Track costs with the savings ledger.

Stay informed about policy changes.
Check ArXiv’s announcements regularly.
Adjust your workflow as new tools and rules appear.

Source: https://the-decoder.com/arxiv-caps-submissions-at-two-per-month-as-ai-paper-flood-overwhelms-the-preprint-server/