1 comments

  • werdnabae 23 minutes ago

    Really interesting work on reproducing reward hacking behaviors across frontier AI models.