---
title: DynamicVLA DOM Demo
emoji: 🦾
colorFrom: green
colorTo: indigo
sdk: gradio
sdk_version: 6.24.0
app_file: app.py
short_description: Predict DynamicVLA action chunks on DOM
python_version: "3.12"
startup_duration_timeout: 1h
---

# DynamicVLA · DOM Demo

Unofficial Hugging Face Space for [`hzxie/dynamic-vla-DOM`](https://huggingface.co/hzxie/dynamic-vla-DOM) —
a 0.4B vision-language-action policy for **dynamic object manipulation**.

Upload wrist + opposite-camera frames and a language instruction. The Space
loads the official checkpoint and returns a **20-step end-effector action chunk**
(xyz + Euler + gripper) with a 3D path plot.

This is a single-chunk policy demo. Closed-loop evaluation still needs the
official [Isaac Lab stack](https://github.com/hzxie/DynamicVLA).

- Paper: [arXiv:2601.22153](https://arxiv.org/abs/2601.22153)
- Dataset: [`hzxie/DOM`](https://huggingface.co/datasets/hzxie/DOM)
- Project: [infinitescript.com/project/dynamic-vla](https://www.infinitescript.com/project/dynamic-vla/)

Weights are released under NTU S-Lab License 1.0. This Space vendors the
policy code from the official repo for inference only.