Add robotics task category, project links and data download instructions

README.md

@@ -1,10 +1,44 @@
 ---
 license: apache-2.0
+task_categories:
+- robotics
 ---
-## Paper
-[arXiv:2503.03480](https://arxiv.org/abs/2503.03480)
 
+# SafeVLA Dataset
+
+This repository contains the data and assets for the paper [SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning](https://arxiv.org/abs/2503.03480), accepted at NeurIPS 2025 (Spotlight).
+
+[Project Page](https://pku-safevla.github.io) | [GitHub Repository](https://github.com/PKU-Alignment/SafeVLA)
+
+## Dataset Description
+SafeVLA provides a framework and data for safety alignment in Vision-Language-Action (VLA) models. The dataset includes:
+- **Optimized Objaverse assets**: Annotations and 3D assets.
+- **ProcTHOR-Objaverse houses**: Procedurally generated environments for robot training and evaluation.
+- **Training Data**: Imitation Learning (IL) data for tasks such as Fetch, Pickup, and Object Navigation.
+
+## Sample Usage
+
+To download the different components of the dataset, you can use the scripts provided in the [official repository](https://github.com/PKU-Alignment/SafeVLA):
+
+### 1. Downloading Objaverse assets and annotations
 ```bash
+python -m objathor.dataset.download_annotations --version 2023_07_28 --path /path/to/objaverse_assets
+python -m objathor.dataset.download_assets --version 2023_07_28 --path /path/to/objaverse_assets
+```
+
+### 2. Downloading ProcTHOR-Objaverse houses
+```bash
+python -m scripts.download_objaverse_houses --save_dir /path/to/objaverse_houses --subset val
+```
+
+### 3. Downloading training data
+```bash
+# TASK_TYPES can be: FetchType | PickupType | ObjectNavType
+python -m scripts.download_training_data --save_dir /path/to/training_data --task_types TASK_TYPES
+```
+
+## Citation
+```bibtex
 @inproceedings{zhang25safevla,
     title={SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning},
     author={Borong Zhang and Yuhao Zhang and Jiaming Ji and Yingshan Lei and Josef Dai and Yuanpei Chen and Yaodong Yang},