GPT3.5 used to hallucinate a lot but it was drastically improved with 4, to a point where I can't even remember the last time it did it. It feels like people ar

GPT3.5 used to hallucinate a lot but it was drastically improved with 4, to a point where I can't even remember the last time it did it. It feels like people are exaggerating or basing it on the experience with old models.