Datasets:

---

zhangzixin02

/

PhysToolBench

like 5

This PR significantly enhances the dataset card for PhysToolBench by:

• Adding task_categories: ['image-text-to-text'] to the metadata for better discoverability, reflecting its nature as a Visual Question Answering (VQA) dataset.
• Including a direct link to the Hugging Face paper page (https://huggingface.co/papers/2510.09507).
• Providing a comprehensive description of the dataset, including its motivation, structure (three difficulty levels), and purpose, derived from the paper's abstract and the GitHub README.
• Embedding the teaser image from the GitHub repository to visually represent the benchmark.
• Incorporating a detailed "Sample Usage" section with instructions and code snippets for environment setup, dataset download, model inference (proprietary and open-source), and evaluation, directly sourced from the official GitHub repository.
• Adding the BibTeX citation for proper attribution and an Acknowledgement section.