LehongWu/opsd-v3_8_2-collect_6tasks_013789_v3_1_0416-gemini3flash_medium-repeat8_0416_all500trajs

该数据集包含图像、学生和教师提示、完成文本以及额外信息，额外信息包括目标、任务特定提示、先前指令、专家推理、专家指令、轨迹键、步骤ID和消息行索引等结构化字段。数据集包含10,000个训练样本。

This dataset includes images, student and teacher prompts, completion texts, and additional structured information. The additional information contains structured fields such as objectives, task-specific prompts, prior instructions, expert reasoning, expert instructions, trajectory keys, step IDs, and message line indices. The dataset consists of 10,000 training samples.