LATIDIA · Ciberseguridad

Respuestas correctas, hallazgos no respaldados: vinculación de evidencia en la reconstrucción forense de registros de agentes de LLM

arXiv:2610.09581v1 Tipo de anuncio: nuevo Resumen: La reconstrucción forense de las acciones del agente LLM requiere no solo recuperar el valor correcto, sino establecer qué registro conservado respalda ese hallazgo. Registros de herramientas, gener

La noticia

arXiv:2610.09581v1 Announce Type: new Abstract: Forensic reconstruction of LLM-agent actions requires not only recovering the correct value, but establishing which preserved record supports that finding. Tool logs, generated explanations, and local citation identifiers capture different parts of this evidence, yet a citation identifier does not establish a source unless its binding to a record is preserved. We audit this distinction using 64 mechanically checkable cases from saved AgentDojo Banking executions. Two LLM readers reconstruct source relationships under controlled variations in visible evidence and identifier-to-record bindings. We separately evaluate complete-record agreement, evidence-grounded findings, justified abstention, and unsupported assertions. With original identifiers and no binding table, Sonnet recovered every literal source location