LATIDIA · Robótica

TILDE: Destilación de profesor en el circuito a partir de modelos de cimentación incorporados

arXiv: 2610.25376v1Tipo de anuncio: nuevo Resumen: La capacitación previa a la acción a gran escala equipa a los Modelos de cimentación incorporados (EFM) con fuertes capacidades de manipulación, con fortalezas particulares en la generalización y el ciclo cerrado r

La noticia

arXiv:2610.25376v1 Announce Type: new Abstract: Large-scale action pretraining equips Embodied Foundation Models (EFMs) with strong manipulation capabilities, with particular strengths in generalization and closed-loop robustness, but reproducing such pretraining requires substantial robot data and computation. To transfer these capabilities from pretrained EFMs to policies without action pretraining, we introduce TILDE (Teacher-In-the-Loop Distillation from EFMs), which synergizes on-policy distillation at student-induced states with learning from teacher-executed recovery trajectories. We further propose AutoDAgger, an autonomous data aggregation pipeline in which a robotic reward model is utilized to monitor task progress and trigger teacher intervention, thereby collecting recovery experience without human monitoring or manual control. AutoDAgger accelerates recovery-trajectory collection by more than $2\times$