LATIDIA · Robótica

Retroceda el mundo, mantenga la reflexión: reflexión inducida por retroceso para agentes de LLM de Long-Horizon

arXiv: 2609.18304v1 Tipo de anuncio: Cross Resumen: Los agentes del modelo de lenguaje grande (LLM) abordan cada vez más las tareas de largo horizonte a través de la interacción del entorno de varios pasos, sin embargo, una sola acción errónea puede alterar el st posterior

La noticia

arXiv:2609.18304v1 Announce Type: cross Abstract: Large language model (LLM) agents increasingly tackle long-horizon tasks through multi-step environment interaction, yet a single erroneous action can alter subsequent states and observations, causing errors to compound over time. Existing methods either correct the context without repairing altered environment states or restore earlier states while discarding useful experience, making it difficult to both eliminate failure conditions and avoid repeating past mistakes. We argue that reliable recovery should instead be treated as a rollback-boundary control problem that jointly determines when to intervene, where to resume, and what information should survive recovery. Based on this view, we propose Rollback-Induced Reflection (RIR), a unified recovery framework that