17 pointsby jessehu4 hours ago1 comment
  • werdnabae4 hours ago
    Really interesting work on reproducing reward hacking behaviors across frontier AI models.