Add paper link and task category to dataset card

This PR adds a link to the paper, the appropriate `task_categories`, and clarifies the dataset description.

README.md

@@ -20,4 +20,10 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+task_categories:
+- image-text-to-text
 ---
+
+This dataset, LongWriter-V-22K, is used to train the LongWriter-V model for ultra-long and high-fidelity generation in vision-language models.  It contains image-text pairs designed for prompting the model to generate extended text descriptions based on the input image.
+
+The model and dataset are described in the paper: [LongWriter-V: Enabling Ultra-Long and High-Fidelity Generation in Vision-Language Models](https://arxiv.org/abs/2502.14834)