GoToLocal

该数据集是一个8x8网格布局的格子世界环境，其中的任务是根据语言指令导航至目标物体。每个轨迹包含一条语言指令、一个状态序列以及一个动作序列。该数据集的规模包括100万个训练轨迹，其任务专注于导航至目标物体。

This dataset is a grid-world environment with an 8x8 grid layout, where the task is to navigate to a target object following natural language instructions. Each trajectory in the dataset includes a natural language instruction, a state sequence, and an action sequence. The dataset contains 1,000,000 training trajectories, with all tasks focused on navigating to target objects.