Festivus

policy

BC-Z

Google / UC Berkeley / Stanford · TensorFlow

or hover any field below to flag it

Overview

Name
BC-Z
Author
Google / UC Berkeley / Stanford
Framework
TensorFlow
License
apache-2.0
Skill type
manipulation
Evidence level
verified
Task description
Zero-shot task generalization via large-scale behavioral cloning. Trained on 25,877 teleoperated demos across 100 tasks (125 hours). Conditioned on language embeddings or human video demos. Generalizes to 24 unseen tasks without task-specific demonstrations.

Spaces

Action space
end-effector-pose · 7-dim · 10Hz
Observation space
  • type: multimodal
  • · primary_rgb
  • · language_embedding

Links

HuggingFace repo
null
Paper (arXiv)
https://arxiv.org/abs/2202.02005

Compatible robots

0

No robots list BC-Z as compatible yet. Know of one? Flag it above.

Compatible environments

2
tabletop-cleannot in seedtabletop-clutterednot in seed

Datasets that reference this policy

0

No datasets reference BC-Z yet.