22f3001825/aegisweave-trajectories

该数据集包含交互式环境中的步骤记录，用于模拟或强化学习任务。特征包括episode_id（剧集ID）、step（步骤编号）、state_text（状态文本描述）、action_id（动作标识符）、action_params（动作参数，如stakeholder_id相关方ID和tone语气）、harmful（是否有害）、action_success（动作是否成功）、schema_failure（模式失败标识）和step_reward（步骤奖励值）。数据集分为训练集，包含10,000个示例，总大小约为6.2MB。

This dataset contains step records from an interactive environment, designed for simulation or reinforcement learning tasks. Features include episode_id, step, state_text (state description), action_id, action_params (with stakeholder_id and tone), harmful (indicating harmfulness), action_success (action success flag), schema_failure (schema failure indicator), and step_reward (reward value). It is split into a training set with 10,000 examples and a total size of approximately 6.2MB.