Datasets:

---

Everloom

/

AAWR-DROID

like 0

This PR significantly improves the dataset card for Everloom/AAWR-DROID by:

• Consolidating Metadata: Moving the existing metadata (license, task_categories, language, tags, and size_categories) from the Markdown content to the correct YAML front matter at the top of the file. The size_categories value has been corrected to 10B<n<100B to accurately reflect the dataset's 13.5GB size.
• Adding Links: Including explicit links to the associated paper (Real-World Reinforcement Learning of Active Perception Behaviors), the official project page (https://penn-pal-lab.github.io/aawr/), and the GitHub repository (https://github.com/penn-pal-lab/aawr) for easy access to related resources.
• Adding Sample Usage: Incorporating a "Sample Usage" section with setup instructions and a code snippet for training AAWR on a toy simulated task, directly extracted from the GitHub README. This provides clear guidance for users on how to get started with the associated code.
• Adding Citation: Including a "Citation" section with the BibTeX entry from the GitHub repository to facilitate proper academic attribution.
• Refining Content: Removing the non-standard "Please comment below with your suggestions" text and organizing the existing "searching behavior" website under an "Additional Resources" section for better structure.