← back to paper
arxiv: 2608.06729 · 2 revisions
AtlasVLA: Persistent World-Ego State Modeling for Vision-Language-Action Models