This PR improves the model card by:
- Adding the `pipeline_tag: image-text-to-text` metadata so that the models are discoverable under the VQA/multimodal category on the Hub.
- Adding a link to the paper [GridVQA-X: A Framework for Evaluating Multimodal Explainability Methods](https://huggingface.co/papers/2606.14740) and its official GitHub repository.
- Including the BibTeX citation for the paper.
