Restoring Linguistic Grounding in VLA Models via Train-Free Attention Recalibration

arXiv · 2026-03-06 · Low-Level Learning-Based Action Modelling / Input Modelling / 2D Vision Language Action Models with Model-agnostic Strategies