fossil-eai/babyai-xgen-bot-random-successful-2

提供的README内容描述了一个与强化学习任务相关的数据集，特别是导航和交互场景。数据集包括任务ID、时间步、动作、图像观察、奖励（折扣和非折扣）以及各种类型的反馈（任务反馈、规则反馈和指令反馈）等特征。数据分为训练集和验证集，并提供了每个分割的字节数和示例数。该数据集似乎用于涉及目标导向行为的任务，具有多种奖励类型和反馈机制。

The provided README content describes a dataset related to reinforcement learning tasks, specifically focusing on navigation and interaction scenarios. The dataset includes features such as task IDs, timesteps, actions, image observations, rewards (both discounted and undiscounted), and various types of feedback (task, rule, and instruction feedback). The data is split into training and validation sets, with detailed information on the number of bytes and examples for each split. The dataset appears to be used for tasks involving goal-directed behavior, with multiple reward types and feedback mechanisms.