Two ML reviewers report fabricated citations or LLM-generated text in 15 of 22 submissions

Sources: primary, discussion
Summary: Two reviewers report fabricated citations or generated text in 15 of the 22 submissions they reviewed. They state that two submissions carrying invented authors substituted onto real papers were accepted as orals, conditional on fixing the references. They released an MIT-licensed bibliography audit that resolves each entry against Crossref, arXiv, DataCite, and Semantic Scholar.
Why it matters: The failure survived review in the reported cases, and the released audit tool makes citation resolution a mechanical check rather than a reviewer judgment call.