← Leaderboard

RT-1-X

VLA
Google DeepMind · Oct 2023 · built on RT-1

RT-1 retrained on the Open X-Embodiment mixture from 22 robots, released as TensorFlow and JAX checkpoints. The paper self-reports about 50% higher success than the original per-dataset methods in small-data domains; it is a standard SimplerEnv baseline, where it scores near zero on the WidowX tasks.

Type
VLA
Parameters
—
Backbone
–
Weights
Open
License
Apache-2.0 (code) / CC-BY-4.0 (other materials)
Commercial use
Allowed
Released
Oct 2023
Robots
Google Robot, WidowX (Bridge), Open X-Embodiment robots
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

1 of 13 benchmarks
SimplerEnv WidowXSaturatedSimulation
#29 of 301.1%, reported by a third party
Not reported on 12 benchmarks
Wrong or missing number? Report an error