@misc{indiciae6bada98dbcce,
  title = {Direction-Scale Decomposition in Action Representation: Rethinking What to Tokenize for Vision-Language-Action Models},
  author = {Yufei Duan and Hang Yin and Alberta Longhini and Chao Tang and Danica Kragic},
  year = {2026},
  url = {https://arxiv.org/abs/2609.28865},
  note = {Source identifier: 2609.28865}
}