HUMANAV

该数据集包含了用于评估不同大型语言模型从非正式驾驶指令生成驾驶逻辑验证（DLV）代码能力的导航指令。此外，数据集还包含了一份驾驶规则手册以及在不同实验中呈现的各种导航任务。规模上，该数据集涉及多个复杂度不同的实验。其任务旨在评估大型语言模型在生成自动驾驶场景下的导航约束能力。

This dataset contains navigation instructions used to evaluate the ability of different large language models to generate driving logic verification (DLV) code from informal driving instructions. In addition, the dataset also includes a driving rule manual and various navigation tasks presented across different experiments. In terms of scale, the dataset involves multiple experiments with varying levels of complexity. The tasks in this dataset aim to evaluate the navigation constraint generation capabilities of large language models in autonomous driving scenarios.