A review of the proceedings from four major computer-science conferences showed that none from 2021, and all from 2025, had fake citations.

[petersuber@fediscience.org]

A review of the proceedings from four major computer-science conferences showed that none from 2021, and all from 2025, had fake citations.
https://arxiv.org/abs/2602.05867v1

The authors prefer the term "mysterious citations" which they define this way: "No paper [with] a similar enough title exists. The cited location either does not exist or holds an unrelated paper with different authors."

#AI #LLMs #Hallucinations #Misconduct #ScholComm

[emmatonkin@mstdn.social]

@petersuber "No author within our dataset acknowledged using AI to generate citations even though all four conference policies required it, indicating current policies are insufficient. "

A very strange mystery indeed. Perhaps we should call Poirot in so he can point out the obvious answers on page 1, knock off work early and spend the rest of the book wandering around delicatessens tasting brands of hot chocolate, or something.

[emmatonkin@mstdn.social]

@petersuber
This whole thing reminds me very much of RFC 3514.

Well over two decades after the Evil Bit was first proposed (a one-bit security flag in IP v4 to indicate that a packet is sent with evil intent), we still see apparently malicious packets arriving at firewalls, yet these do *not* have the Evil Bit set. However, as the Evil Bit is not set, we cannot say with certainty whether these apparently malicious packets are, in fact, evil.