Just thought I'd highlight this issue to the ML community, since I recently had this problem arise and it might be useful for some. I had a coauthor who I knew was somewhat untrustworthy when it came to LLM use. This coauthor added some last-minute new references to the paper. The deadline was near,
