← Leaderboard

V-JEPA 2-AC

World-action
Meta · Jun 11, 2025

Latent action-conditioned world model post-trained from the self-supervised V-JEPA 2 video encoder on under 62 hours of DROID video, used for planning toward image goals. Results are self-reported zero-shot on Franka arms in two labs (e.g. 80% cup and 50% box pick-and-place), not on a public benchmark.

Type
World-action
Parameters
—
Backbone
V-JEPA 2 ViT-g (1B) encoder + 300M action-conditioned predictor
Weights
Open
License
MIT
Commercial use
Allowed
Released
Jun 11, 2025
Robots
Franka (DROID)
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

Not reported on 13 benchmarks
Wrong or missing number? Report an error