neomatrix369/py-bug-trace-laguna-xs-2-l2-rollouts
收藏资源简介:
该数据集包含15个训练样本,用于对话系统的训练。每个样本包括提示(prompt)和完成(completion)的多轮对话,其中提示和完成都包含角色(role)和内容(content)字段,完成还包含推理内容(reasoning_content)。此外,还提供了元信息(如ID、难度、类别)、奖励分数、时间戳(包括设置、生成、评分、模型和环境时间)、完成状态、截断状态、停止条件、指标(如精确匹配奖励和轮数)以及令牌使用统计。数据集旨在支持对话生成或强化学习任务。
This dataset contains 15 training examples for dialogue system training. Each example includes a multi-turn prompt and completion, where both prompt and completion contain role and content fields, and the completion additionally includes reasoning_content. It also provides meta-information (such as ID, difficulty, category), reward scores, timing information (including setup, generation, scoring, model, and environment timing), completion status, truncation status, stop conditions, metrics (such as exact_match_reward and num_turns), and token usage statistics. The dataset is intended to support dialogue generation or reinforcement learning tasks.




