Universal and Transferable Adversarial Attacks on Aligned Language Models Explained

แชร์
ฝัง
  • เผยแพร่เมื่อ 7 ต.ค. 2024
  • Paper found here: arxiv.org/abs/...
    Demo here: llm-attacks.org/

ความคิดเห็น • 10

  • @machinelearnear
    @machinelearnear ปีที่แล้ว +1

    Amazing work, Gabriel, really like your videos & approach. Keep it up!

  • @amirhoseinshirani7328
    @amirhoseinshirani7328 ปีที่แล้ว

    Your approach to present articles is awesome! keep working 👏🏻

  • @viruldewnaka1193
    @viruldewnaka1193 ปีที่แล้ว

    Great stuff, keep uploading

  • @SkyBeast55
    @SkyBeast55 ปีที่แล้ว

    i like your videos! thanks a lot

  • @AndyLee-xq8wq
    @AndyLee-xq8wq 10 หลายเดือนก่อน

    nice video!!

  • @jimmyjackson7848
    @jimmyjackson7848 7 หลายเดือนก่อน

    reading over these papers reminds me of Johnny Long, when he introduced the google hacking for pen-testers..

  • @shivangitripathi1356
    @shivangitripathi1356 10 หลายเดือนก่อน

    how are tokens generated actually? how do we check attacks takes place by placing those tokens? from where does these token comes? can anybody answer me ?

  • @noadsensehere9195
    @noadsensehere9195 4 หลายเดือนก่อน

    How can I implement this paper

  • @רותםישראלי-כ3ד
    @רותםישראלי-כ3ד ปีที่แล้ว

    Really liked your videos but I prefer the ones about vision

    • @gabrielmongaras
      @gabrielmongaras  ปีที่แล้ว

      Glad you're enjoying my videos! I'm trying to keep a wide range of topics covering vision, text, and audio as there are really cool developments in all three domains! Trying not to get trapped in a single domain as developments in one domain can also affect another. Also, LLMs are the craze right now and I think reading over some of the papers would be beneficial to know what's currently going on with them right now.