brains/act · model record
ACT (Action Chunking with Transformers)
Conditional-VAE transformer policy that predicts chunks of future joint actions, introduced with the low-cost ALOHA bimanual teleoperation system. Learns fine manipulation from about 50 demonstrations per task.
01
Embodiments
robots it has been shown on02
Papers naming this model
from the papers feed- Mind2Act World: A Diagnostic Benchmark for Embodied Mind, Physical Action, and Mind-Action Composition in Robotic Manipulation
- Body Schema Policy: Body as the Vision-Action Interface for Efficient Robot Loco-Manipulation
- ROSE: A Shared Cross-Embodiment Facial-Motion Latent Space for Humanoid Robot Expression