Add paper link and GitHub link, specify task category and license

README.md

@@ -1,4 +1,5 @@
 ---
+pretty_name: J
 dataset_info:
   features:
   - name: json_schema
@@ -61,12 +62,15 @@ configs:
     path: data/JsonSchemaStore-*
   - split: Glaiveai2K
     path: data/Glaiveai2K-*
-pretty_name: J
+license: mit
+task_categories:
+- text-generation
 ---
 
 # JSONSchemaBench
 
-[![Paper](https://img.shields.io/badge/Paper-arXiv-blue)](https://arxiv.org/abs/2501.10868)  
+[![Paper](https://img.shields.io/badge/Paper-arXiv-blue)](https://arxiv.org/abs/2501.10868)
+[![GitHub](https://img.shields.io/badge/Code-GitHub-blue)](https://github.com/guidance-ai/jsonschemabench)
 
 JSONSchemaBench is a benchmark of **real-world JSON schemas** designed to evaluate **structured output generation** for Large Language Models (LLMs). It contains approximately **10,000 JSON schemas**, capturing diverse constraints and complexities.
 
@@ -108,7 +112,6 @@ Each dataset split contains:
 🚀 **For more details, check out the [paper](https://arxiv.org/abs/2501.10868).**
 
 ## 📚 Citation
-If you use this dataset, please cite:
 ```bibtex
 @misc{geng2025jsonschemabench,
       title={Generating Structured Outputs from Language Models: Benchmark and Studies},