Paper
LLMs & Texto
What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics
Jailbreak attacks expose vulnerabilities in aligned large language models, revealing that harmful intent is encoded in structured intermediate uncertainty dynamics rather than outp…
Hugging Face · Daily Papers
·Sofiia Nikolenko, Michele Papucci
·
·▲ 3 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Sofiia Nikolenko, Michele Papucci, Mina Rezaei, Shireen Kudukkil Manchingal
- 3 upvotes da comunidade
- Temas: Large Language Models, jailbreak attacks, safety training, token-level predictive entropy, logit lens, prompt-level entropy
Resumo
Resumo original (em inglês), extraído do paper:
Jailbreak attacks expose vulnerabilities in aligned large language models, revealing that harmful intent is encoded in structured intermediate uncertainty dynamics rather than output representations.
