Update dataset card with paper link, abstract, task category, tags, and sample usage

#2
by nielsr HF Staff - opened

This PR significantly enhances the nilshoehing/rocketsciencebench dataset card by:

  • Adding task_categories and tags to metadata:
    • task_categories: image-text-to-text accurately reflects the dataset's focus on spatial relation understanding in Vision Language Models.
    • tags: spatial-reasoning, vlm, benchmark improve discoverability for users looking for similar resources.
  • Updating the Dataset Description: Replaced the placeholder with the comprehensive abstract from the paper "Understanding Space Is Rocket Science - Only Top Reasoning Models Can Solve Spatial Understanding Tasks" to provide a detailed overview of the dataset.
  • Updating the Paper Link: Replaced the "coming soon" placeholder with the official Hugging Face paper URL: https://huggingface.co/papers/2509.02175.
  • Adding a Sample Usage section: Included a basic code snippet demonstrating how to load the dataset using the datasets library, based on the information in the Github README.
  • Adding a BibTeX Citation: Provided a BibTeX entry for the paper to facilitate proper academic attribution.

These changes make the dataset card more informative, discoverable, and user-friendly.

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment