MBZUAI-Paris/EgyptianBench

该数据集包含了四种类型的数据配置：长句单轮翻译、短句单轮翻译、文本摘要和文本转写。每种类型的数据都包含了唯一的标识符、数据集名称、翻译方向（文本摘要没有翻译方向）以及消息内容，其中消息内容又分为内容和角色两个部分。每种配置都提供了测试集分割，包括字节数和示例数。

The dataset includes four types of data configurations: single-turn long sentence translation, single-turn short sentence translation, text summarization, and text transliteration. Each type of data contains a unique identifier, dataset name, translation direction (text summarization does not have a translation direction), and message content, which is divided into content and role. Each configuration provides a test split, including the number of bytes and the number of examples.