Comment by cainxinth

Not a paper, but a startup called Vectara claimed to be investigating LLM hallucination/ confabulation rates last year:

https://www.nytimes.com/2023/11/06/technology/chatbots-hallu...