Improve dataset card: add metadata, license, and links

README.md

@@ -1,10 +1,19 @@
 ---
-license: mit
+license: cc-by-4.0
+task_categories:
+- robotics
+tags:
+- vision-language-action
+- vla
+- benchmark
+- robot-learning
 ---
 
 # Dataset Card for LIBERO-PRO Perturbation Dataset
 
-This dataset contains the **`bddl`** and **`init`** files of LIBERO-PRO configurations under **object**, **relation**, **semantic**, **task**, and **environment** perturbations. The dataset supports direct integration with the [LIBERO-PRO](https://github.com/Zxy-MLlab/LIBERO-PRO) framework.
+[**Project Page**](https://zxy-mllab.github.io/LIBERO-PRO-Webpage/) | [**Paper**](https://huggingface.co/papers/2510.03827) | [**Code**](https://github.com/Zxy-MLlab/LIBERO-PRO)
+
+This dataset contains the **`bddl`** and **`init`** files of LIBERO-PRO configurations under **object**, **relation**, **semantic**, **task**, and **environment** perturbations. The dataset supports direct integration with the [LIBERO-PRO](https://github.com/Zxy-MLlab/LIBERO-PRO) framework to evaluate Vision-Language-Action (VLA) models beyond rote memorization.
 
 ---
 
@@ -15,7 +24,7 @@ This dataset contains the **`bddl`** and **`init`** files of LIBERO-PRO configur
 - **Curated by:** LIBERO-PRO Research Team  
 - **Affiliation:** MLLab, Huazhong University of Science and Technology  
 - **Language(s) (NLP):** English (instructional text)  
-- **License:** MIT  
+- **License:** CC-BY-4.0 (for dataset artifacts), MIT (for codebase)
 - **Primary Purpose:** Evaluation of VLA models under structured perturbations  
 
 This dataset extends the original [LIBERO benchmark](https://github.com/Lifelong-Robot-Learning/LIBERO/) by introducing **systematic perturbations** in five dimensions:
@@ -33,11 +42,11 @@ Each perturbation includes corresponding **`init` files** (initial environment c
 
 **How to use:**
 1. Copy all **`.bddl`** files to:
-   ```
+   ```bash
    LIBERO-PRO/libero/libero/bddl_files/
    ```
 2. Copy all **`init`** files to:
-   ```
+   ```bash
    LIBERO-PRO/libero/libero/init_files/
    ```
 3. Follow the quick start instructions provided in the [LIBERO-PRO README](https://github.com/Zxy-MLlab/LIBERO-PRO#readme).
@@ -54,15 +63,22 @@ Each perturbation category contains:
 
 ## Citation
 
-If you use this dataset, please cite the LIBERO-PRO project:
+If you use this dataset, please cite both the original LIBERO benchmark and the LIBERO-PRO project:
 
 **BibTeX:**
 ```bibtex
-@misc{zhou2025liberopro,
-  title={LIBERO-PRO: Benchmarking Perturbation Robustness of Visual Language Action Models},
-  author={Zhou, Xueyang and Tie, Guiyao and Zhang, Guowen and Wang, Hechang and Zhou, Pan},
-  year={2025},
-  howpublished={\url{https://github.com/Zxy-MLlab/LIBERO-PRO}},
+@article{zhou2025liberopro,
+  title={LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization},
+  author={Xueyang Zhou and Yangming Xu and Guiyao Tie and Yongchao Chen and Guowen Zhang and Duanfeng Chu and Pan Zhou and Lichao Sun},
+  journal={arXiv preprint arXiv:2510.03827},
+  year={2025}
+}
+
+@article{liu2023libero,
+  title={LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning},
+  author={Liu, Bo and Zhu, Yifeng and Gao, Chongkai and Feng, Yihao and Liu, Qiang and Zhu, Yuke and Stone, Peter},
+  journal={arXiv preprint arXiv:2306.03310},
+  year={2023}
 }
 ```
 
@@ -78,4 +94,4 @@ If you use this dataset, please cite the LIBERO-PRO project:
 ## Dataset Card Contact
 
 For questions or issues, please contact:  
-📧 **d202480819@hust.edu.cn**
+📧 **d202480819@hust.edu.cn**