LATIDIA · Ciberseguridad

Ataques de decodificación controlada en Black-Box LLM

arXiv:2609.36956v1 Announce Type: new Abstract: Manipulating next-token probabilities during generation can bypass the safety alignment of large language models. Sin embargo, los enfoques existentes se basan en el acceso al modelo Weig

La noticia

arXiv:2609.36956v1 Announce Type: new Abstract: Manipulating next-token probabilities during generation can bypass the safety alignment of large language models. Existing approaches, however, rely on access to model weights or numerical token probabilities and therefore do not apply to interfaces that return only sampled text. Reconstructing probabilities from sampled outputs offers a possible alternative, but finite sampling produces sparse and noisy estimates, while repeating this process at every generation step incurs substantial query costs. Our empirical observations suggest that large distributional changes along successful jailbreak trajectories are concentrated at a small subset of positions, motivating selective control. We introduce \method{}, a framework for jailbreaking through text-only continuation interfaces that permit repeated