home.social

#jailbreakdetection — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #jailbreakdetection, aggregated by home.social.

fetched live
  1. This paper introduces a model-agnostic threat evaluation using N-gram language models to measure jailbreak likelihood, finding discrete optimization attacks more effective than LLM-based ones and that jailbreaks often exploit rare bigrams.

    Read more: arxiv.org/abs/2410.16222

    #AIResearch #JailbreakDetection

  2. This paper introduces a model-agnostic threat evaluation using N-gram language models to measure jailbreak likelihood, finding discrete optimization attacks more effective than LLM-based ones and that jailbreaks often exploit rare bigrams.

    Read more: arxiv.org/abs/2410.16222

    #AIResearch #JailbreakDetection