This PR improves the dataset card for SpaceNum by:
- Adding the `image-text-to-text` task category and relevant tags.
- Linking to the official paper: [SPACENUM: Revisiting Spatial Numerical Understanding in VLMs](https://huggingface.co/papers/2605.23898).
- Providing a link to the project page.
- Adding a brief description of the SpaceNum framework and its two primary tasks (Num2Space and Space2Num) based on the paper abstract.
