Audit finds 68% of sampled ML conference submissions contain fabricated or AI-generated content

A July 30 audit of 22 submissions to three ML conferences found 68% had fabricated citations, fake author lists, or clearly LLM-generated text. Of 11 papers checked for citations, 55% had fabricated ones. Of 11 checked for author or LLM issues, 82% were flagged. The sample was small and non-random. The author released a bib-audit tool on GitHub.