← Leaderboard

MiniVLA

VLA
Stanford · Dec 9, 2024

1B-parameter VLA from Stanford ILIAD (Belkhale and Sadigh) built in the OpenVLA codebase. It replaces Llama 2 7B with a Qwen2.5-0.5B Prismatic VLM and adds residual-VQ action chunking and multi-image inputs. The released checkpoints are trained on LIBERO-90 and Bridge V2.

Type
VLA
Parameters
1B
Backbone
Prismatic (Qwen2.5-0.5B, DINOv2 + SigLIP)
Weights
Open
License
MIT
Commercial use
Allowed
Released
Dec 9, 2024
Robots
Franka Panda (sim), WidowX (Bridge)
SO-101
–
LeRobot
–
Inputs
RGB · L
VRAM
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

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