LATIDIA · Investigación

ARIA: Inspección y modificación en tiempo de ejecución para sistemas multiagente

arXiv:2610.20780v1 Tipo de anuncio: nuevo Resumen: En los sistemas multiagente (MAS) basados en LLM, un error local puede propagarse a medida que otros agentes se basan en él, convirtiendo un error corregible en una tarea fallida. Diagnóstico de ejecuciones completadas

La noticia

arXiv:2610.20780v1 Announce Type: new Abstract: In LLM-based multi-agent systems (MAS), a local mistake can propagate as other agents build on it, turning a correctable error into a failed task. Diagnosing completed runs or repeatedly re-executing them to search for corrections does not provide a learned policy for intervening while execution is still in progress. We present Agent Runtime Inspection and Amendment (ARIA), a framework for learning when and how to repair ongoing multi-agent executions. Our key intuition is to turn execution-verified repairs into reusable supervision, moving repair search into data construction and training so that a learned model can intervene at runtime without repeatedly testing alternatives. First, we collect over