Large language models often know when they are being evaluated

(arxiv.org)

56 points | by jonbaer 11 hours ago ago

75 comments