Add paper link, task category and code link

README.md

@@ -9,9 +9,15 @@ dataset_info:
     num_examples: 27332
   download_size: 487651276
   dataset_size: 1076886272
+task_categories:
+- text-generation
 configs:
 - config_name: default
   data_files:
   - split: train
     path: data/train-*
 ---
+
+This dataset contains data used in the paper [From Hours to Minutes: Lossless Acceleration of Ultra Long Sequence Generation up to 100K Tokens](https://huggingface.co/papers/2502.18890).
+
+Code: https://github.com/bigai-nlco/TokenSwift