LATIDIA · Ciberseguridad

Red-TTT: Capacitación a tiempo de prueba para modelos automatizados de lenguaje grande para el desguace de cárceles

arXiv:2610.05282v1 Tipo de anuncio: Cross Resumen: Los modelos de lenguaje grandes siguen siendo vulnerables a los jailbreaks, y el equipo rojo automatizado es la forma estándar de encontrar jailbreaks en modelos de lenguaje grandes a escala. Métodos actuales

La noticia

arXiv:2610.05282v1 Announce Type: cross Abstract: Large language models remain vulnerable to jailbreaks, and automated red teaming is the standard way to find jailbreaks in large language models at scale. Current methods either draw more samples at test time through search, rewriting, and tree expansion, or train a stronger attacker offline with reinforcement learning. Both share a limitation: once an attack on a specific target behavior begins, the attacker's weights are frozen. Any signal it gathers about the behavior stays in its context window and is discarded afterward. The attacker never adapts its proposal distribution mid-attack, so success depends almost entirely on the sampling budget, and under a budget affordable at