[Crash0v3rid3]

> GPT-4 doesn't hallucinate all that much.

What data do you have to back this up?

From my own experience GPT-4 hallucinates quite a bit, enough to make it unusable for my use cases.

[ToValueFunfetti]

The technical report[1] makes that claim at least:

>GPT-4 significantly reduces hallucinations relative to previous GPT-3.5 models (which have them- selves been improving with continued iteration). GPT-4 scores 19 percentage points higher than our latest GPT-3.5 on our internal, adversarially-designed factuality evaluations

[1] https://arxiv.org/abs/2303.08774 (text from page 10)

[simonw]

"reduces hallucinations" and "doesn't hallucinate all that much" aren't quite the same thing.

[ToValueFunfetti]

I interpreted "all that much" as "close to as much as the earlier model", but yours is probably a more fair reading.