Post

While I generally agree with this policy, the focus on hallucinated citations specifically makes me uncomfortable: I usually get my BibTeX from Google Scholar. Yes, GS is sometimes wrong. No, I don't carefully check my references. Is using AI to fill in references really morally any different?

This policy as stated wouldn’t ban you over a dead link, but you should check (all of) your references enough to know that they exist. Or do you have a different failure mode in mind?

Here's an example of a "hallucinated" citation from a decade ago (i.e. pre-LLMs). The same bad citation appeared in multiple papers. I eventually traced the source to Google Scholar (and it's now fixed).

Not "by those authors." The point Thomas is making (which tbh I find quite minor, but it's a point nonetheless) is that there is *no* such real paper by an author named "Marcus Hardt". The author name is wrong. Would that be flagged by the arXiv's soon-to-be checking system?

My point is that references *specifically* are an odd thing to focus on. Some people are pedantic about refs, but many (most?) are just copying BibTeX from various imperfect websites without checking. Yes, there are different degrees of bad refs, but I don't see a clear moral distinction being made.

I’ve refereed numerous papers now w LLM errors in refs that are so much worse. Made up papers that don’t exist at all by any authors. Papers w a completely incorrect set of authors (not just typos), papers that exist but don’t say anything like what they’re cited for (or even say the opposite!)