Improve dataset card: Add task category, library name, tags, and sample usage

This PR enhances the dataset card for LiveMCPBench by:
- Adding `task_categories: text-generation` to the metadata, as specified, and `library_name: datasets` for clarity on how to programmatically access the dataset.
- Including relevant `tags` such as `llm-agents`, `tool-use`, `benchmark`, and `mcp` for improved discoverability on the Hugging Face Hub.
- Expanding the "Dataset Description" section with more details from the paper's abstract, providing a comprehensive overview of the benchmark's scope and components.
- Adding a "Dataset Structure" section to briefly describe the dataset's file organization.
- Introducing a "Sample Usage" section with a Python snippet to demonstrate how to load and inspect the dataset using the Hugging Face `datasets` library.

These updates aim to make the dataset card more informative and user-friendly.

README.md

@@ -1,15 +1,23 @@
 ---
-license: apache-2.0
 language:
 - en
-pretty_name: LiveMCPBench
+license: apache-2.0
 size_categories:
 - n<1K
+pretty_name: LiveMCPBench
+task_categories:
+- text-generation
+library_name: datasets
+tags:
+- llm-agents
+- tool-use
+- benchmark
+- mcp
 configs:
 - config_name: default
   data_files:
   - split: test
-    path: "tasks/tasks.json"
+    path: tasks/tasks.json
 ---
 
 <a id="readme-top"></a>
@@ -34,9 +42,27 @@ configs:
 </p>
 
 ## Dataset Description
-LiveMCPBench is a benchmark for evaluating the ability of agents to navigate and utilize a large-scale MCP toolset. 
+LiveMCPBench is the first comprehensive benchmark designed to evaluate LLM agents at scale across diverse Model Context Protocol (MCP) servers. It comprises 95 real-world tasks grounded in the MCP ecosystem, challenging agents to effectively use various tools in daily scenarios within complex, tool-rich, and dynamic environments. To support scalable and reproducible evaluation, LiveMCPBench is complemented by LiveMCPTool (a diverse collection of 70 MCP servers and 527 tools) and LiveMCPEval (an LLM-as-a-Judge framework for automated and adaptive evaluation). The benchmark offers a unified framework for benchmarking LLM agents in realistic, tool-rich, and dynamic MCP environments, laying a solid foundation for scalable and reproducible research on agent capabilities.
+
+## Dataset Structure
+The dataset consists of `tasks.json`, which contains the 95 real-world tasks used for benchmarking LLM agents.
+
+## Sample Usage
+
+You can load the dataset using the Hugging Face `datasets` library:
+
+```python
+from datasets import load_dataset
+
+# Load the dataset
+dataset = load_dataset("ICIP/LiveMCPBench")
+
+# Print the dataset structure
+print(dataset)
 
-It provides a comprehensive set of tasks that challenge agents to effectively use various tools in daily scenarios.
+# Access an example from the test split
+print(dataset["test"][0])
+```
 
 ## Citation
 
@@ -51,3 +77,4 @@ If you find this project helpful, please use the following to cite it:
       primaryClass={cs.AI},
       url={https://arxiv.org/abs/2508.01780}, 
 }
+```