Agent-STAR-TravelDataset

This repository contains the synthetic datasets for the paper Demystifying Reinforcement Learning for Long-Horizon Tool-Using Agents: A Comprehensive Recipe.

Official GitHub Repository: WxxShirley/Agent-STAR

Dataset Description

The Agent-STAR TravelDataset provides over 17K synthetic queries designed for the TravelPlanner testbed. TravelPlanner is a long-horizon tool-use environment where agents must iteratively call tools to satisfy multifaceted constraints. This dataset supports the STAR [Data Synthesis → SFT → RL] pipeline presented in the paper.

Dataset Files

Usage

To generate your own training samples or run inference using these datasets, please follow the instructions and environment setup provided in the official GitHub repository. Note that you will also need the Travel Database (CSV files) for environment interaction.

Citation

@misc{wu2026agentstar,
      title={Demystifying Reinforcement Learning for Long-Horizon Tool-Using Agents: A Comprehensive Recipe}, 
      author={Xixi Wu and Qianguo Sun and Ruiyang Zhang and Chao Song and Junlong Wu and Yiyan Qi and Hong Cheng},
      year={2026},
      eprint={2603.21972},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2603.21972}, 
}