fossil-eai/babyai-xgen-bot-successful

该数据集包含多个配置，每个配置包含与任务执行、动作、图像观察、奖励和各种反馈机制相关的特征。数据集包括预处理和原始数据格式，具有任务ID、时间步长、动作、图像观察、终止状态、不同类型的奖励（折扣、非折扣、标准化）以及任务、规则、任务反馈和指令的嵌入等特征。数据集分为训练集和验证集，每个分割的字节数和示例数都有详细统计。该数据集似乎与任务或任务执行环境相关，可能涉及强化学习或类似的AI训练场景。

The dataset includes multiple configurations, each containing features related to task execution, actions, image observations, rewards, and various feedback mechanisms. The dataset comes in both preprocessed and raw data formats, with features such as task_id, timesteps, actions, img_obss (image observations), terminations, different types of rewards (discounted, undiscounted, normalized), and embeddings for missions, rules, tasks, and instructions. The dataset is divided into training and validation sets, with detailed statistics on the number of bytes and examples for each split. The dataset appears to be related to a task or mission execution environment, possibly involving reinforcement learning or similar AI training scenarios.