Recently, the well-known preprint platform arXiv announced an update to its review rules for papers in the Computer Science (CS) category, in response to the rapidly increasing number of review and position papers generated or assisted by AI, which has become difficult to manage.

New Rules: Review and Position Papers in CS Must Be Peer-Reviewed First

According to the new rules, Review Papers and Position Papers submitted to the arXiv CS category must be formally accepted and completed with peer review before they can be uploaded.

When submitting these two types of papers, authors must attach a proof of successful peer review (such as journal references and DOI metadata). Review and position papers without such proof are highly likely to be rejected by arXiv.

AI robot writing a paper

AI Exacerbates the Proliferation of Papers, Especially in the CS Category

arXiv admitted that this adjustment is due to the fact that the platform has been "overwhelmed by papers" in recent years, and the development of AI has exacerbated this problem. AI makes it fast and easy to write papers that do not introduce new research findings. Although the number of submissions in all categories is growing, the situation is particularly noticeable in the arXiv CS category.

arXiv pointed out that the platform now receives hundreds of review papers each month, most of which are simply "annotated lists of references" without substantial discussion on open research issues.

Official Response: Not a Policy Change, but a Reinforcement and Implementation

In its official statement, arXiv emphasized that strictly speaking, this is not a change in submission rules. The platform stated that review papers and position papers were never considered acceptable paper types.

In the past, arXiv only accepted these two types of papers at the discretion of reviewers, when the number was small and the quality was extremely high (for example, those written by senior researchers or scientific associations), because they could generate widespread interest. However, now, facing the surge of low-quality AI-assisted content, relying on external peer review has become the only way to filter truly valuable papers.

Objective: Free Up Volunteers and Accelerate the Dissemination of High-Quality Results

arXiv made this change for two main objectives:

  1. Improve Quality: Help readers find review papers and position papers written by subject experts that truly have academic value.

  2. Reduce the Load: Allow volunteer reviewers to focus on the types of papers that the platform accepts, reduce the delay caused by other papers, and thus enable scientific discoveries to be disseminated faster.

Notably, papers whose research topic is "the impact of technology on society" and are classified under cs.CY or physics.soc-ph can still be submitted without peer review. For other paper categories, arXiv has left room for flexibility. The platform stated that if similar situations occur in other categories in the future, similar review rules may be adopted, and announcements will be made separately.