ArXiv versus AI slop: will capping preprint submissions work?

The preprint server arXiv has implemented a new policy limiting researchers to submitting two papers per month, starting October 1, to manage the overwhelming number of manuscripts it receives. This decision comes in response to a significant increase in submissions, largely attributed to content generated by artificial intelligence tools. In September alone, 40,000 preprints were uploaded, compared to 30,000 in previous months. Steinn Sigurðsson, arXiv’s scientific director, notes that the cap aims to prevent the site’s volunteer moderators from being inundated, as they currently review about 1,000 manuscripts daily. While some researchers, like Siddhesh Zādéy, find the limit reasonable, others, such as Victor Galitski, believe it could be stricter. Similar measures are being adopted by other platforms, like SocArXiv, which limits submissions to one article per author per week. However, some argue that simply rejecting more articles won’t solve the problem, as these papers may just be submitted elsewhere. QUESTION: How might the rise of AI-generated content impact the quality and credibility of scientific research in the future?