Update task category to image-text-to-text

This PR updates the `task_categories` metadata tag from `text-generation` to `image-text-to-text`. This change better reflects the dataset's nature as a comprehensive benchmark for evaluating LLM agents that navigate and interact with a diverse set of Model Context Protocol (MCP) tools in real-world, potentially multimodal, environments, as described in the paper "LiveMCPBench: Can Agents Navigate an Ocean of MCP Tools?".

README.md

@@ -4,9 +4,9 @@ language:
 license: apache-2.0
 size_categories:
 - n<1K
-pretty_name: LiveMCPBench
 task_categories:
-- text-generation
+- image-text-to-text
+pretty_name: LiveMCPBench
 library_name: datasets
 tags:
 - llm-agents
@@ -42,7 +42,7 @@ configs:
 </p>
 
 ## Dataset Description
-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.
+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 that enables 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.