2 comments

  • Retro_Dev 1 hour ago
    > Overall, our results indicate that current language models possess some functional introspective awareness of their own internal states. We stress that in today’s models, this capacity is highly unreliable and context-dependent; however, it may continue to develop with further improvements to model capabilities.
    • Escapade5160 9 minutes ago
      If it's context dependent is it really introspection?
  • whythismatters 1 hour ago
    >Submitted on 5 Jan 2026