17 points | by jessehu 7 hours ago
1 comments
Really interesting work on reproducing reward hacking behaviors across frontier AI models.
Really interesting work on reproducing reward hacking behaviors across frontier AI models.