zzwkk/

MUA-RL-Dataset

Paper | Github

This dataset is part of the MUA-RL (MULTI-TURN USER-INTERACTING AGENT REINFORCEMENT LEARNING FOR AGENTIC TOOL USE). The dataset contains 1,580 carefully annotated tasks composed of high-quality dialogue conversations and associated tool sets.

The dataset consists of two main components:

5 Mock Tasks: Simulated conversational scenarios 4 MCP (Model Context Protocol) Tasks: Real-world… See the full description on the dataset page: https://huggingface.co/datasets/zzwkk/MUA-RL-Dataset.