Datasets:

---

Junteng

/

Vision4Chart

like 0

This PR enhances the dataset card for the "On the Perception Bottleneck of VLMs for Chart Understanding" dataset by:

• Adding task_categories: ['image-text-to-text'] and language: en to the metadata for better categorization.
• Including relevant tags such as chart-understanding, vlm, multimodal, and clip to improve discoverability.
• Adding a direct link to the associated GitHub repository (https://github.com/hkust-nlp/Vision4Chart) in the content section.

These additions will make the dataset card more informative and easier to find on the Hugging Face Hub.