LATIDIA · Investigación

Recuperabilidad como primitiva del sistema para agentes de IA de Long-Horizon

arXiv: 2609.13672v1Tipo de anuncio: nuevo Resumen: los agentes de IA pueden ser interrumpidos mientras editan archivos, llaman a herramientas o llevan a cabo tareas de varios pasos. El reinicio repite el trabajo completado, pero continúa desde no verificado o superado

La noticia

arXiv:2609.13672v1 Announce Type: new Abstract: AI agents can be interrupted while editing files, calling tools, or carrying out multi-step tasks. Restarting repeats completed work, but continuing from unverified or outdated progress can carry earlier errors forward. A saved state is not necessarily a suitable place to resume. We introduce recoverability as a system primitive that makes reuse an explicit decision: select a supported starting point and a permitted recovery action, or withhold automatic continuation. Its behavioral contract binds that choice to supporting evidence, execution, and independent checks. A reference architecture connects persistence, validation, and control, with complementary runtime instances testing distinct responsibilities. Four deterministic and 20 paired file challenges demonstrate