How about getting credit for research you never did nor actually exisits? Hello 2024 in AI years. Laura Czerniewicz writes of getting an Scholar Alert for a paper she never authored:
It could be from a hallucinated paper or just an LLM query that returned something that looks “research-y” but is not. The thing is our intuition seems to operate from a place that LLMs are looking directly at “real” sources, when they are not. They do not draw directly from them, its only the patters of words in authors or titles or an abstract that spin out a suppose citation.
I’d say that LLMs hallucinate as much when they get a citation correct- its generated the same way as a faux one. They don’t trudge off to the library to check their work, as loong as it appears like a citation, its all the same. So at some level, its rather impressive when you get an accurate reference, right? That’s quite a feat.
There are of course worrisome things like fake journals out there. For some fun, see Tyler Vigen’s Spurious Scholar which generates fake papers generated via random correlations between variables in a data set.
See Winds and Webs: An Exploration of the Correlation Between Internet Sites and Wind Power Worldwide (The Journal of Renewable Energy Riddles (2024) 47, 52-63) published by “Elserver”
And also, on the fake generated paper front, have a play with MathGen “a program to randomly generate professional-looking mathematics papers, including theorems, proofs, equations, discussion, and references.” created by mathematician Nate Eldredge.
How are you thinking about references to research papers “out there”?