← Leaderboard

LAPA

VLA
KAIST AI · Oct 15, 2024

Latent Action Pretraining from Videos (KAIST with the University of Washington, Microsoft Research, NVIDIA and Ai2): a VLA pretrained to predict quantized latent actions learned from video without action labels, then fine-tuned on a small labeled set to output robot actions. Its SimplerEnv WidowX number comes from fine-tuning on 100 rollouts collected in the simulator, not from the standard zero-shot protocol.

Type
VLA
Parameters
7B
Backbone
Large World Model (LWM) 7B
Weights
Open
License
MIT
Commercial use
Allowed
Released
Oct 15, 2024
Robots
WidowX (sim), Franka, Language Table (sim)
SO-101
–
LeRobot
–
Inputs
RGB · L
VRAM
–

Benchmark results

1 of 13 benchmarks
SimplerEnv WidowXSaturatedSimulation
  • 57.3%, self-reported by the authors, not a leaderboard entryPaper / model card ↗Oct 2024· fine-tuned on 100 SIMPLER rollouts· non-standard
57.3%, self-reported by the authors, not a leaderboard entrynot ranked
Not reported on 12 benchmarks
Wrong or missing number? Report an error