ChinaTravel
收藏资源简介:
ChinaTravel是由南京大学国家重点实验室开发的一个真实世界基准数据集,专门用于评估语言代理在中国旅行规划中的应用。该数据集涵盖了中国10个最受欢迎城市的旅行信息,包括720个航班和5770趟列车,以及3413个景点、4655家餐厅和4124家酒店的详细信息。数据集通过问卷调查收集用户需求,并设计了一个可扩展的领域特定语言来支持自动评估。ChinaTravel旨在解决复杂的真实世界旅行规划问题,特别是在多兴趣点行程安排和用户偏好满足方面,为语言代理在旅行规划中的应用提供了重要的测试平台。
ChinaTravel is a real-world benchmark dataset developed by the State Key Laboratory of Nanjing University, specifically designed to evaluate the application of language agents in travel planning within China. This dataset covers travel information for 10 of China's most popular cities, including 720 flights, 5770 train trips, as well as detailed information of 3413 tourist attractions, 4655 restaurants and 4124 hotels. User requirements are collected via questionnaire surveys, and a scalable domain-specific language is designed to support automatic evaluation. ChinaTravel aims to address complex real-world travel planning problems, especially in itinerary arrangement with multiple points of interest and meeting user preferences, providing an important testbed for the application of language agents in travel planning.
ChinaTravel: A Real-World Benchmark for Language Agents in Chinese Travel Planning
数据集概述
- 名称: ChinaTravel
- 描述: 一个用于中文旅行规划的语言代理的现实世界基准。
作者
- Jie-Jing Shao<sup>1</sup>
- Xiao-Wen Yang<sup>1</sup>
- Bo-Wen Zhang<sup>1</sup>
- Bai-Zhi Chen
- Wen-Da Wei
- Lan-Zhe Guo
- Yu-Feng Li
机构
- 南京大学 LAMDA 小组
相关链接
- arXiv: https://arxiv.org/abs/2412.13682
- 代码: https://github.com/LAMDASZ-ML/ChinaTravel (即将推出)

- 1ChinaTravel: A Real-World Benchmark for Language Agents in Chinese Travel Planning南京大学 · 2024年



