Computer science conferences are using different caps to manage submissions and reviewer workloads, but the available figures do not establish that generative AI caused the surge. Organizers cite very high submission volumes, limited review capacity, and concerns about low-quality or AI-generated work; the rules they have adopted vary by conference and track.
Why are computer science conferences limiting submissions?
Conference organizers need enough qualified reviewers to assess each paper. When submissions outgrow that capacity, organizers face pressure on review quality, workloads, and the fairness of the process. Recent policies address that problem in different ways: some cap the number of papers naming an author, while others constrain reciprocal-review assignments.
AAAI’s August 25, 2025 update reported almost 29,000 submissions to the AAAI-26 Main Technical Track. After removing submissions that failed to comply with policies—including the author cap—roughly 23,000 remained under review, nearly twice the number of papers reviewed by AAAI-25. AAAI also reported that almost 20,000 submissions came from China. These are AAAI’s figures for that cycle, not an independently audited count or a census of the field.
Organizers have also raised concerns about low-quality or AI-generated submissions. That concern is not the same as evidence that AI caused the overall increase: the conference-specific figures do not isolate AI’s contribution to submission growth.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitches#1 Best Overall
What do the current caps require?
The rules differ in what they limit, who they apply to, and what happens when submissions exceed a ceiling. The figures and requirements below apply to the named conference edition or track.
| Conference or track | What is capped | Limit and enforcement | AI-use rule described by organizers |
|---|---|---|---|
| ICLR 2027 | Author-listed submissions, with a separate rule for new authors | ICLR Program Chairs’ September 2, 2026 post sets a 20-paper limit for all authors and a one-paper limit for “new authors”—people who have never published at a major AI, machine-learning, computer-vision, robotics, or NLP conference. The post discusses rate-limiting submissions; it does not specify a surplus-rejection mechanism in the information reported here. | Organizers cite low-quality and AI-generated submissions as concerns. The stated concern is not, by itself, a blanket rule against AI assistance. |
| AAAI-26 Main Technical Track | Author-listed submissions | AAAI’s August 25, 2025 update says submissions exceeding its cap were among those removed for noncompliance. The numeric ceiling is not stated in that update. | The cited update discusses integrity measures, but does not establish a general AI-writing allowance or prohibition for this track. |
| OSDI ’26 | Author-listed submissions | USENIX’s call for the 2026 edition limits submissions to eight per author. The cited call does not state a specific surplus-rejection procedure. | Work wholly or largely generated by AI is treated as fraud; AI editing tools are acceptable. This policy applies to OSDI ’26, not automatically to other USENIX events. |
| SIGCOMM ’26 | Author-listed submissions | The ACM SIGCOMM call says that if an author appears on more than eight submissions, surplus submissions will be desk-rejected. | The call points authors to ACM’s generative-AI policy, including disclosure of AI-created content. |
| ICML 2026 | Reciprocal-review assignments | The call caps the number of papers that can designate the same person as a reciprocal reviewer; a numeric ceiling is not stated in the information reported here. | Generative AI tools, including large language models, may assist with writing or research. |
| NeurIPS 2026 Position Paper Track | AI use in paper preparation, rather than an author-count cap described here | A numeric submission cap is not stated in the track’s June 2, 2026 statement. | Papers must be substantially human-written; AI use is restricted to copy-editing or similar peripheral changes. This is a track-specific rule, not a universal NeurIPS policy. |
How are author caps different from reciprocal-review limits?
An author cap limits how many submissions may list a person as an author. It can reduce the volume associated with prolific contributors, but a paper may have many coauthors, and an author ceiling can affect research groups differently. ICLR’s Program Chairs said they found it difficult to imagine one author contributing meaningfully to 20 papers and that a substantially smaller ceiling could have been defensible. They also acknowledged the risk of placing an unfair burden on junior researchers in large groups.
A reciprocal-review limit instead constrains how many papers can assign the same person as a reviewer in return for submitting. It targets a review-capacity mechanism rather than a person’s total authorship count. ICML 2026 describes this kind of cap, while also permitting generative AI tools to assist writing or research—two separate policy choices within one conference cycle.
ICLR’s Program Chairs reported that about 20% of ICLR 2026 submissions had no reciprocal reviewer. They calculated this by comparing submission author lists with the final program-committee list. That figure is specific to their method and that conference cycle; it should not be treated as a field-wide rate.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Did AI cause the surge in conference papers?
The available conference reports do not establish that. AAAI’s submission count and ICLR’s discussion of AI-generated or low-quality content show that organizers are responding to scale and raising integrity concerns. They do not measure what share of increased submissions was generated or enabled by AI, nor do they separate AI’s effect from other possible reasons for higher volume.
AI rules also vary by venue and track. OSDI ’26 draws a line between work largely generated by AI and AI editing; SIGCOMM ’26 points to ACM disclosure requirements; ICML 2026 permits AI assistance for writing or research; and the NeurIPS 2026 Position Paper Track restricts AI to peripheral changes. These examples cannot be generalized into a single policy for computer science conferences.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Do submission caps make peer review fairer?
They are intended to address capacity and related concerns, but the policy examples do not demonstrate that caps have reduced reviewer hours, improved review quality, or made outcomes fairer. A 2025 arXiv paper proposes a framework for fairness-aware desk rejection and discusses the difficulty of choosing which papers to reject when limits are exceeded. It is an analytical proposal, not evidence that existing caps have achieved those results.
For authors, the practical step is to read the current call for the exact conference and track, check whether its ceiling counts all coauthored papers or only reciprocal-review assignments, and verify the stated consequence for exceeding it. Do not assume that an allowance or restriction at one venue applies to another.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
