← back to paper
arxiv: 2606.07100 · 2 revisions
LARA: Latent Action Representation Alignment for Vision-Language-Action Models