NATURAL PLAN
收藏资源简介:
NATURAL PLAN是由Google DeepMind创建的一个自然语言规划基准数据集,包含三个主要任务:旅行规划、会议规划和日历调度。该数据集通过使用Google Flights、Google Maps和Google Calendar等工具的真实数据,构建了多种约束条件,以评估大型语言模型(LLMs)在自然语言规划任务中的表现。数据集的创建过程涉及随机生成城市、时间和地点等参数,并添加特定约束以确保任务的唯一解。NATURAL PLAN的应用领域主要集中在提升LLMs在复杂任务规划中的能力,特别是在处理多城市、多人和多日程的复杂场景中。
NATURAL PLAN is a natural language planning benchmark dataset created by Google DeepMind. It comprises three core tasks: travel planning, meeting planning, and calendar scheduling. The dataset builds diverse constraint conditions using real-world data from tools including Google Flights, Google Maps, and Google Calendar, to evaluate the performance of Large Language Models (LLMs) in natural language planning tasks. The dataset creation process involves randomly generating parameters such as cities, times and locations, and adding specific constraints to ensure a unique solution for each task. The application scope of NATURAL PLAN primarily focuses on enhancing the capabilities of LLMs in complex task planning, especially when handling complex scenarios involving multiple cities, multiple participants and multiple schedules.

- 1NATURAL PLAN: Benchmarking LLMs on Natural Language PlanningGoogle DeepMind · 2024年



