OpenHelix-Team/VLA-Adapter

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2294 stars · 208 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

VLA-Adapter is the official implementation of a tiny-scale vision-language-action model that bridges vision-language representations to robot actions without large backbones or extensive pre-training. It includes checkpoints, LIBERO benchmark models, and real-world ALOHA deployment support.

Use cases

• train a small vision-language-action model for robot manipulation
• deploy a VLA model on ALOHA hardware
• evaluate robot policies on the LIBERO benchmark
• run robot control inference without a 7B backbone
• fine-tune a lightweight VLA model for a robotic arm

When to choose

• you need efficient robot manipulation policies without large-scale pre-training
• you want state-of-the-art LIBERO results with a tiny model
• you have limited GPU resources for VLA training

When to avoid

• you need a general-purpose large VLA backbone
• your robot hardware is not yet among the supported platforms
• you need non-manipulation embodied tasks

Facets

library · maturity active

machine-learning deep-learning robotics robotics machine-learning artificial-intelligence python vision-language-action embodied-ai robot-manipulation aloha libero tiny-scale-models linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="OpenHelix-Team/VLA-Adapter")