On The Truthfulness of 'Surprisingly Likely' Responses of Large Language Models

Open in new window