LATIDIA · Investigación

INDAGAR: Aprender a auditar el cumplimiento de la política en el uso de herramientas de agentes de LLM

arXiv: 2610.22124v1Tipo de anuncio: nuevo Resumen: Los sistemas agénticos modernos amplían los modelos de lenguaje con herramientas externas, lo que les permite completar tareas complejas a través de interacciones de varias vueltas. A medida que estos agentes se despliegan en

La noticia

arXiv:2610.22124v1 Announce Type: new Abstract: Modern agentic systems extend language models with external tools, enabling them to complete complex tasks through multi-turn interactions. As these agents are deployed in consequential settings, they are often constrained by policies governing pricing, privacy, confirmation, safety, and other requirements. However, existing evaluation methods leave a critical gap: capability benchmarks measure what agents can accomplish, while content-safety red-teaming measures harmful generations, but neither tests whether agents comply with predefined policies during otherwise valid tool use. Such policy violations are difficult to audit because their effects may emerge only after several turns, span multiple tool calls, or surface in downstream systems. We propose PROBE (Policy Rule