brains/act · model record

ACT (Action Chunking with Transformers)

Name: ACT (Action Chunking with Transformers)
Author: Stanford University

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.

Company

Stanford University

Kind

VLA

Access

Open code

License

MIT

Released

23 Apr 2023

Embodiments

robots it has been shown on

Stanford ALOHAstanford/aloha

Papers naming this model

from the papers feed

No paper in the current feed names ACT (Action Chunking with Transformers) in its title.