Study finds LLMs can identify their own mistakes

Robot peering inside head of another robot


It turns out that LLMs encode quite a bit of knowledge about the truthfulness of their answers, even when they give the wrong one.Read More