← back to paper
arxiv: 2605.11459 · 2 revisions
Overcoming Dynamics-Blindness: Training-Free Pace-and-Path Correction for VLA Models