LATIDIA · Ciberseguridad
Reducción de la diversidad de atacantes en el «red teaming» adaptativo de LLM
arXiv:2611.19680v1 Tipo de anuncio: nuevo Resumen: El aprendizaje por refuerzo puede hacer que los «red teaming» automatizados tengan más éxito y, al mismo tiempo, reducir los ataques que generan. Entrenamiento conjunto de un atacante y un defensor con políticas relativas al grupo
WhatsApp ↗Telegram ↗
La noticia
arXiv:2611.19680v1 Announce Type: new Abstract: Reinforcement learning can make automated red-teaming more successful while narrowing the attacks it produces. Co-training an attacker and a defender with group-relative policy optimization (GRPO) has been reported to reduce attacker diversity. This experiment reproduces attack diversity reduction using three runs with a Qwen2.5-14B attacker and a Phi-4-mini defender trained by direct preference optimization (DPO). The attacker receives a success-only reward and uses a length-unbiased loss, bounded updates, warm-up and ratio clipping. The attacker is evaluated on 26 AdvBench behaviors. Training ended after 12 rounds, following a decision based on development checkpoints, so the reported intervals are nominal. Attack Diversity Retention (ADR) compares the within-behavior