Improve dataset card: Add metadata, paper and code links

This PR improves the dataset card for `PAI-Bench-Transfer` by:
- Adding `license: mit` to the YAML metadata, as indicated in the GitHub repository.
- Adding `task_categories: ['video-to-video']` to the YAML metadata, reflecting the dataset's focus on video-based tasks.
- Including a direct link to the paper ([https://huggingface.co/papers/2512.01989](https://huggingface.co/papers/2512.01989)) at the top of the card for easy access.
- Adding a direct link to the GitHub repository ([https://github.com/SHI-Labs/physical-ai-bench](https://github.com/SHI-Labs/physical-ai-bench)) at the top of the card for easy access.

These changes enhance the dataset card's completeness and discoverability on the Hugging Face Hub.

README.md

@@ -1,16 +1,21 @@
 ---
-configs:
-  - config_name: benchmark
-    data_files:
-      - split: PAIBenchTransfer
-        path: metadata.csv
 language:
-  - en
+- en
 pretty_name: PAI-Bench-Transfer
+configs:
+- config_name: benchmark
+  data_files:
+  - split: PAIBenchTransfer
+    path: metadata.csv
+task_categories:
+- video-to-video
+license: mit
 ---
 
 # Physical AI Bench - Transfer
 
+[Paper](https://huggingface.co/papers/2512.01989) | [Code](https://github.com/SHI-Labs/physical-ai-bench)
+
 This dataset (Phsical AI benchmark, PAI-Bench) consisting of 600 examples across three key scenarios: robotic arm operations, driving, and ego-centric everyday life scenes, each representing a critical aspect of Physical AI. This dataset is constructed by sampling a number of videos from three different datasets. The specific details are provided below.
 
 | Dataset                                                      | Category           | Sample Nums |
@@ -54,4 +59,4 @@ If you use Physical AI Bench in your research, please cite:
       primaryClass={cs.CV},
       url={https://arxiv.org/abs/2512.01989}, 
 }
-```
+```