Path Planning from Natural Language (PPNL)
收藏资源简介:
PPNL数据集由乔治梅森大学计算机科学系创建,用于评估大型语言模型在自然语言路径规划任务中的空间-时间推理能力。数据集包含多种网格环境,大小从5x5到7x7不等,障碍数量从1到11个。该数据集通过随机生成初始和目标位置,以及障碍物的布局,来测试模型在不同环境下的路径规划能力。此外,数据集还包括了不可达目标的情景,以评估模型识别此类情况的能力。PPNL数据集的应用领域主要集中在人工智能和机器人路径规划,旨在解决复杂环境下的智能导航问题。
The PPNL dataset was developed by the Department of Computer Science at George Mason University to evaluate the spatio-temporal reasoning capabilities of large language models (LLMs) in natural language path planning tasks. The dataset includes various grid environments with sizes ranging from 5x5 to 7x7, and the number of obstacles varies between 1 and 11. It tests the path planning performance of models across different environments by randomly generating initial and target positions as well as obstacle layouts. Additionally, the dataset incorporates scenarios with unreachable target positions to assess the model's ability to recognize such situations. The application fields of the PPNL dataset mainly focus on artificial intelligence and robotic path planning, aiming to address intelligent navigation challenges in complex environments.

- 1Can Large Language Models be Good Path Planners? A Benchmark and Investigation on Spatial-temporal Reasoning乔治梅森大学计算机科学系 · 2024年



