LATIDIA · Ciberseguridad

JailBound: FOL-Guided Jailbreak Evaluation for Exploring LLM Behavioral Boundaries

arXiv:2611.07530v1 Anuncio Tipo: nuevo Resumen: A medida que los grandes modelos de lenguaje (LLM) se implementan cada vez más como asistentes conversacionales de propósito general, deben manejar diversas solicitudes de los usuarios mientras rechazan de manera confiable los daños.

La noticia

arXiv:2611.07530v1 Announce Type: new Abstract: As large language models (LLMs) are increasingly deployed as general-purpose conversational assistants, they must handle diverse user requests while reliably refusing harmful ones. Jailbreak attacks craft adversarial instructions or conversational contexts to elicit unsafe responses despite these safeguards. However, existing jailbreak benchmarks often rely on manually designed or collected prompts, making it difficult to systematically extend coverage to new threat conditions. Moreover, existing attack search methods typically optimize for attack success rate (ASR), providing limited insight into how model behavior changes locally between safe refusal and unsafe response. To address these limitations, we present JailBound, a unified jailbreak evaluation framework that combines automated benchmark construction