LATIDIA · Robótica

ESCALA: abstracción de primer concepto de modelo pequeño con internalización de experiencia guiada por LLM

arXiv: 2611.12120v1Tipo de anuncio: cruz Resumen: Los agentes autónomos deben adquirir continuamente conocimientos reutilizables a partir de la experiencia sin depender del costoso razonamiento del modelo de lenguaje grande (LLM) para cada decisión. Sin embargo,

La noticia

arXiv:2611.12120v1 Announce Type: cross Abstract: Autonomous agents must continually acquire reusable knowledge from experience without relying on costly large language model (LLM) reasoning for every decision. However, existing experience-based approaches often struggle to transform accumulated observations and reasoning into compact knowledge that lightweight models can independently apply. This paper proposes SCALE, a small-model-first framework that integrates selective LLM reasoning, experience-driven concept abstraction, and knowledge internalization. A lightweight model handles routine predictions, while prediction errors and accumulated cross-episode evidence selectively trigger LLM reflection. The LLM analyzes relevant historical experiences, generates positive and negative counterfactual examples, and proposes reusable concept hypotheses. Independently validated concepts are progressively incorporated into hierarchical neural representations, enabling