X-VLA
- AI model
- Humanoids and robot learning
X-VLA is a vision-language-action model that uses soft prompts to train one Transformer backbone across many robot embodiments, with a server-client setup for inference.
Source on GitHub: github.com/2toinf/X-VLA (opens in a new tab). Facts read at commit 6bc2513 (opens in a new tab).
Licenses
- Code
- Apache-2.0 (opens in a new tab)
Read by Robotheca from LICENSE (opens in a new tab) (Apache-2.0) at commit 6bc2513 (opens in a new tab), on 16 Oct 2026.
- Commercial use
- Yes
What it does
- Uses embodiment-specific learnable soft prompts to guide a unified Transformer toward multi-domain robot policy learning.
- The X-VLA-0.9B architecture is reported to generalize across six simulation platforms and three real-world robots.
- Runs as an HTTP inference server exposing a POST /act endpoint that takes proprioception, images and a language instruction.
- Provides pretrained foundation and fine-tuned checkpoints for benchmarks such as LIBERO, CALVIN, RoboTwin2 and Simpler, plus LoRA adapters.
- Includes a training script that fine-tunes on custom demonstrations through a modular dataset handler interface.
- Uses a standardized 20-dimensional EE6D control space for actions across different embodiments.
Written by Robotheca from the project's README (opens in a new tab) at commit 6bc2513.
Getting started
- Difficulty
- Advanced
- How to start
- The project's website (opens in a new tab)
Maintenance
- Status
- Maintained: a commit in the last 12 months
- Newest commit
- 27 Jan 2026 on main
- Latest release
- None published on GitHub
Checked on GitHub on 16 Oct 2026.
Links
Robots on Robotheca that use it
No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.
Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.