ArXiv’s LLM Dilemma: Policy Shifts and the Gatekeeping of AI-Generated Scholarship

arXiv just reset the rules for AI-era preprints. The CS category now requires that reviews and position papers be accepted at a journal or conference and complete peer review before submission, a move prompted by the rise of LLMs. This isn’t a ban on preprints—it's gatekeeping designed to curb low-effort AI-generated content ^[1].

What changed at arXiv Before submission to arXiv’s CS category, review articles and position papers must be accepted at a journal or conference and undergo peer review. The policy tweak is explicit in arXiv’s latest update ^[1].

Community voices on gatekeeping and openness - Stopgap against LLM spam — ArXiv CS will no longer accept literature reviews, surveys or position papers without peer-review, a policy shift tied to concerns about LLM-generated content ^[2]. - Tension between openness and credibility — some see gatekeeping as preserving quality, while others worry it narrows open dissemination and makes arXiv less of a free preprint hub ^[2]. - 'Paper dump' fears and the future of AI literature — commenters discuss the risk of labeling quick AI summaries as legitimate scholarship and debate where AI-generated work should land ^[2].

Impact on dissemination and credibility The changes push AI research toward peer-reviewed venues, potentially boosting credibility but narrowing the immediacy of preprint-style sharing ^[1]. While some praise the guardrails, others warn of stifling openness and rapid iteration in a fast-moving field ^[2].

Closing thought: the coming months will reveal whether this balance helps AI scholarship scale responsibly or slows its public, iterative nature.

Referenced posts: ^[1], ^[2]

References