osunlp/TravelPlanner

TravelPlanner是一个用于评估语言代理在工具使用和复杂规划中能力的基准数据集，特别是在多约束条件下的旅行规划。数据集包含1,225个查询，分为训练集、验证集和测试集。训练集包含5个查询及其对应的人类注释计划，验证集包含180个查询，测试集包含1,000个查询。每个查询涉及交通、餐饮、景点和住宿等多方面的规划。

TravelPlanner is a benchmark dataset for evaluating the capabilities of language agents in tool usage and complex planning, with a specific focus on travel planning under multiple constraints. The dataset consists of 1,225 queries, which are split into training, validation, and test subsets. The training subset contains 5 queries along with their corresponding human-annotated plans, the validation subset includes 180 queries, and the test subset has 1,000 queries. Each query involves multi-faceted planning covering transportation, dining, tourist attractions, and accommodation.

原始信息汇总

数据集概述

名称: TravelPlanner

目的: 用于评估语言代理在工具使用和多约束条件下的复杂规划能力。

详细信息: 具体细节可参考相关论文。