TY  - RPRT
TI  - Fast Enough to Act: Spatio-Temporal Visual Token Merging for Low-Latency Robotic VLMs and VLAs
AU  - Junzhou Chen
AU  - Jindong Wang
AU  - Gang Zhou
PY  - 2026
UR  - https://arxiv.org/abs/2606.29350
ID  - 2606.29350
ER  -