Add robotics metadata, paper links, and usage information

README.md

@@ -1,29 +1,67 @@
+---
+license: apache-2.0
+pipeline_tag: robotics
+library_name: lerobot
+tags:
+- vision-language-action
+- vla
+- robot-learning
+---
+
 # A Pragmatic VLA Foundation Model
+
 <p align="center">
-  <img src="assets/Teaser.png" width="100%">
+  <img src="https://github.com/robbyant/lingbot-vla/raw/main/assets/Teaser.png" width="100%">
 </p>
 
+**LingBot-VLA** is a pragmatic Vision-Language-Action (VLA) foundation model designed for robotic manipulation. It is pre-trained on 20,000 hours of real-world data from 9 popular dual-arm robot configurations, achieving strong generalization across various robotic platforms.
+
+- **Paper:** [A Pragmatic VLA Foundation Model](https://huggingface.co/papers/2601.18692)
+- **Repository:** [https://github.com/robbyant/lingbot-vla](https://github.com/robbyant/lingbot-vla)
+- **Project Page:** [https://technology.robbyant.com/lingbot-vla](https://technology.robbyant.com/lingbot-vla)
+
+## Key Features
+- **Large-scale Pre-training**: Trained on 20,000 hours of real-world data.
+- **Strong Performance**: Achieves state-of-the-art results on simulation and real-world benchmarks (GM-100 and RoboTwin 2.0).
+- **Efficiency**: Offers a 1.5 ~ 2.8× speedup in training throughput compared to existing VLA frameworks.
+
+## Related Models
 
-**LingBot-VLA** has focused on **Pragmatic**:
-- **Large-scale Pre-training Data**: 20,000 hours of real-world
-data from 9 popular dual-arm robot configurations.
-- **Strong Performance**: Achieve clear superiority over competitors on simulation and real-world benchmarks.
-- **Training Efficiency**: Represent a 1.5 ∼ 2.8× (depending on the relied VLM base model) speedup over existing VLA-oriented codebases.
+| Model Name | Huggingface | Description |
+| :--- | :---: | :---: |
+| LingBot-VLA-4B &nbsp; | [🤗 lingbot-vla-4b](https://huggingface.co/robbyant/lingbot-vla-4b) | LingBot-VLA *w/o* Depth|
+| LingBot-VLA-4B-Depth | [🤗 lingbot-vla-4b-depth](https://huggingface.co/robbyant/lingbot-vla-4b-depth) | LingBot-VLA *w/* Depth |
 
 ---
 
-## Model Sources
-- Repository: https://github.com/robbyant/lingbot-vla
-- Paper: A Pragmatic VLA Foundation Model
-- Project Page: https://technology.robbyant.com/lingbot-vla
+## Installation
 
-## Related Models
+```bash
+# Install LeRobot
+pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu128
+git clone https://github.com/huggingface/lerobot.git
+cd lerobot
+git checkout 0cf864870cf29f4738d3ade893e6fd13fbd7cdb5
+pip install -e .
+
+# Clone the LingBot-VLA repository
+git clone https://github.com/robbyant/lingbot-vla.git
+cd lingbot-vla/
+pip install -e .
+pip install -r requirements.txt
+```
+
+## Sample Usage
 
-| Model Name | Huggingface | ModelScope | Description |
-| :--- | :---: | :---: | :---: |
-| LingBot-VLA-4B &nbsp; | [🤗 lingbot-vla-4b](https://huggingface.co/robbyant/lingbot-vla-4b) | [🤖 lingbot-vla-4b](https://modelscope.cn/models/Robbyant/lingbot-vla-4b) | LingBot-VLA *w/o* Depth|
-| LingBot-VLA-4B-Depth | [🤗 lingbot-vla-4b-depth](https://huggingface.co/robbyant/lingbot-vla-4b-depth) | [🤖 lingbot-vla-4b-depth](https://modelscope.cn/models/Robbyant/lingbot-vla-4b-depth) | LingBot-VLA *w/* Depth |
+To evaluate the policy on the RoboTwin 2.0 benchmark:
 
+```bash
+export QWEN25_PATH=path_to_Qwen2.5-VL-3B-Instruct
+python -m deploy.lingbot_robotwin_policy \
+ --model_path robbyant/lingbot-vla-4b \
+ --use_length 50 \
+ --port 8080
+```
 
 ---
 
@@ -37,10 +75,5 @@ data from 9 popular dual-arm robot configurations.
 }
 ```
 
----
-
-## License Agreement
-This project is licensed under the [Apache-2.0 License](LICENSE).
-
 ## Acknowledgement
-This codebase is builded on the [VeOmni](https://arxiv.org/abs/2508.02317) project. Thanks for their excellent work!
+This codebase is built on the [VeOmni](https://arxiv.org/abs/2508.02317) project and benefits from [LeRobot](https://github.com/huggingface/lerobot).