LangAGI-Lab/magpie-reasoning-v1-20k-math-verifiable-step-by-step-rationale

该数据集是一个包含多个字段的结构化数据集，主要用于训练自然语言处理模型。它包含了对话数据、生成文本的配置信息、任务相关的元数据以及用于评估和改进模型性能的特征。具体字段包括但不限于：唯一标识符、指令、响应、对话的双方及内容、生成输入和输出的配置、意图、知识类型、难度等级、输入质量、质量说明、任务类别、语言类型、索引、API响应、摘要、演示UUID列表、预测步骤、摘要列表、逐步推理详细说明、答案等。数据集分为训练集，大小为368MB。

The dataset is a structured dataset with multiple fields designed for training natural language processing models. It includes conversational data, configuration information for generating text, metadata related to tasks, and features for evaluating and improving model performance. Specific fields include but are not limited to: unique identifiers, instructions, responses, conversation participants and content, input and output generation configurations, intent, knowledge type, difficulty level, input quality, quality explanation, task categories, language type, index, API responses, summaries, demonstration UUIDs, summary step predictions, summary lists, step-by-step extended rationale, answers, etc. The dataset is split into a training set, which is 368MB in size.