brains/univla · model record
UniVLA
7B VLA from OpenDriveLab and HKU that learns task-centric latent actions from video, including human video without action labels, on a Prismatic backbone. Latent-action planning is then decoded to robot actions for each embodiment.
01
Embodiments
robots it has been shown onNo embodiment is listed in a source yet.
02
Papers naming this model
from the papers feedNo paper in the current feed names UniVLA in its title.