This PR improves the dataset card by:
- Changing `task_categories` from `video-classification` to `other`, since the dataset is designed for audio-visual synchronization reward modeling with human ratings.
- Adding the paper link: https://huggingface.co/papers/2610.16338
- Adding the code repository link: https://github.com/WeChatCV/SyncReward
- Adding the BibTeX citation from the GitHub README.
No other content changes were made.