Best-of-N Jailbreaking
arxiv.org
Best-of-N Jailbreaking
1–10 of 16 posts
Re: Best-of-N Jailbreaking
#2i've never seen such a complicated author list as far as "equal contribution" and "equal advising"
Re: Best-of-N Jailbreaking
#3i've never seen such a complicated author list as far as "equal contribution" and "equal advising"
best of n paper
Re: Best-of-N Jailbreaking
#4It reminds me of that Apple paper. They found that minor changes in the prompt can have large changes in the result.
Re: Best-of-N Jailbreaking
#5Anyone tried this technique against “Gandalf”?
In particular for level 8.
Re: Best-of-N Jailbreaking
#6Never, have I ever read a more complicated abstract.
Re: Best-of-N Jailbreaking
#7Can this be called a "brute force" attack in layman's terms?
Re: Best-of-N Jailbreaking
#8Never, have I ever read a more complicated abstract.
It... seems pretty ordinary to me? Like there isn't even much jargon being used. Try reading a paper in basically any field of hard science!
Re: Best-of-N Jailbreaking
#9Anyone tried this technique against “Gandalf”? https://gandalf.lakera.ai/ In particular for level 8.
There is no level 8... Was this a trick to get me to play it?