Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
I'm not aware of a comprehensive hallucination benchmark, at least one that has been updated for recent SOTA models. If there was, I'd reference it, but hallucination rates have dropped drastically since the 3.5 days (something like 40% of its citations were hallucinate).
I almost never run into them, though I only check important claims. With something like GPT-5T, I'd estimate it's correct north of 95% of the time on factual questions, though I'm not sure if that means 96% or 99.9%.
Uh.. I don't think anything I've said should be interpreted as "they don't happen". Right now, they're uncommon enough that I think you should check only claims that matter, not the exact amount of salt to put in your soup.
More options
Context Copy link