latency-sensitive-bench/demon_attack_fix_latency_4

该数据集包含交互决策任务的数据，用于强化学习或模仿学习应用。每个示例包括图像、分割信息、提示文本、动作ID、动作文本描述、延迟时间、奖励值、完成标志、剧集索引、时间步、随机种子和确定性标志。数据集分为训练集（263,581个示例）和验证集（64,664个示例），总大小约292MB，适用于训练和评估智能体在视觉环境中的决策能力。This dataset contains data for interactive decision-making tasks, suitable for reinforcement learning or imitation learning applications. Each example includes an image, split information, prompt text, action ID, action text description, latency, reward value, done flag, episode index, time step, random seed, and deterministic flag. The dataset is divided into a training set (263,581 examples) and a validation set (64,664 examples), with a total size of approximately 292MB, designed for training and evaluating agent decision-making capabilities in visual environments.