mapauk/mtrag-synthetic-dialogs-annotated-2026-04-27

这是一个对话或问答数据集，包含问题ID、来源、提供者、模型、对话轮次、声明和标注信息。标注涉及对话和轮次级别的评估，如助手理解失败、用户矛盾等，用于分析模型生成内容的质量。数据集包括训练集，共16个示例。

This is a dialogue or question-answering dataset that encompasses question ID, source, provider, model, dialogue turn, statement and annotation information. These annotations cover evaluations at both dialogue and turn levels, including scenarios such as assistant understanding failure and user contradiction, and are used to analyze the quality of model-generated content. The dataset includes a training set with a total of 16 examples.