latency-sensitive-bench/demon_attack_fix_latency_12

该数据集是一个用于强化学习或交互式任务的数据集，包含从模拟或真实环境中收集的轨迹数据。每个示例包括图像数据（image）、分割信息（split）、提示文本（prompt）、动作ID（action_id）、动作文本（action_text）、延迟（latency和latency_ms）、奖励（reward）、完成标志（done）、剧集索引（episode_idx）、时间步（t）、种子（seed）和确定性标志（deterministic）。数据集分为训练集和验证集，训练集有169,767个示例，验证集有44,527个示例，总大小约为190.7MB，适用于训练和评估强化学习模型。This dataset is designed for reinforcement learning or interactive tasks, containing trajectory data collected from simulated or real-world environments. Each example includes image data (image), split information (split), prompt text (prompt), action ID (action_id), action text (action_text), latency (latency and latency_ms), reward (reward), done flag (done), episode index (episode_idx), time step (t), seed (seed), and deterministic flag (deterministic). The dataset is divided into training and validation splits, with 169,767 examples in the training set and 44,527 examples in the validation set, totaling approximately 190.7MB in size, suitable for training and evaluating reinforcement learning models.