chakrabamartya/test_dataset

T1: A Tool-Oriented Conversational Dataset for Multi-Turn Agentic Planning

T1 is a tool-augmented, multi-domain, multi-turn conversational dataset specifically designed to capture and manage inter-tool dependencies across diverse domains.

Overview

We propose T1 dataset (T1) dataset to evaluate agentic workflow applications on their ability to plan and reason in complex, tool-dependent scenarios based on nine distinct domains (4 single-domain and 5 multi-domain)

The listed domains are as follows: Single-domain:

• flight
• hotel
• restaurant
• attraction

Multi-domain:

• flight-hotel
• hotel-restaurant
• hotel-attraction
• flight-hotel-restaurant
• flight-hotel-attraction

T1 consists of a total of 13,500 complex dialogues between a user and an assistant which requires calls to several tools. We have broken down this dataset into train, validation and test partitions.

Contact

E-mail: Amartya Chakraborty, Genta Indra Winata, Nadia Bathaee, Paresh Dashore, Anmol Jain,

Citation

If you find this dataset useful, please cite the following works

@article{amartya2025t1,
  title={T1: A Tool-Oriented Conversational Dataset for Multi-Turn Agentic Planning},
  author={Chakraborty, Amartya and Dashore, Paresh and Bathaee, Nadia and Jain, Anmol and Das, Anirban and Zhang, Shi-Xiong and Sahu, Sambit and Naphade, Milind and Winata, Genta Indra},
  journal={Advances in Neural Information Processing Systems},
  year={2025}
}