← Leaderboard

MoDE

VLA
KIT · Dec 17, 2024

Diffusion transformer policy from KIT (with MIT) that routes tokens to experts by noise level. It has about 436M parameters, excluding the language encoder, and one variant is pretrained on an Open X-Embodiment subset. Its CALVIN and SimplerEnv numbers are self-reported.

Type
VLA
Parameters
—
Backbone
Mixture-of-denoising-experts diffusion transformer
Weights
Open
License
MIT
Commercial use
Allowed
Released
Dec 17, 2024
Robots
Franka Panda (sim), WidowX (sim), Google Robot (sim)
SO-101
–
LeRobot
–
Inputs
RGB · L
VRAM
–

Benchmark results

2 of 13 benchmarks
CALVIN ABC→DNear saturationSimulation
#16 of 244.01 / 5, self-reported by the authors, not a leaderboard entry
SimplerEnv WidowXSaturatedSimulation
#28 of 307.3%, self-reported by the authors, not a leaderboard entry
Not reported on 11 benchmarks
Wrong or missing number? Report an error