This PR improves the dataset card metadata by:
- Adding the relevant `task_categories` field: `automatic-speech-recognition`.
- Removing the `arxiv:2610.08585` tag from the YAML metadata section, since arXiv IDs should only appear in the markdown content (they already appear in the paper link and citation).
No content changes are needed beyond the metadata updates.
