lighteval/synthetic_reasoning

该数据集包含三种配置：归纳(induction)、模式匹配(pattern_match)和变量替换(variable_substitution)。每种配置都有关于特征（源字符串和目标字符串）的信息，以及训练集、测试集和验证集的划分信息，包括每个划分中的示例数量和数据集的字节大小。数据集为每种配置和划分提供了特定的数据文件，并给出了每个配置的总下载大小和完整数据集大小。基于这些信息，该数据集似乎是专注于文本数据，包含源字符串和目标字符串，可能用于各种自然语言处理任务，如文本生成或机器翻译。

The dataset consists of three configurations: induction, pattern_match, and variable_substitution. Each configuration includes information about features (source and target strings), and the splits (train, test, validation) with the number of examples and dataset size in bytes. The dataset is structured with specific data files for each split and configuration, providing the total download size and the full dataset size for each configuration. This dataset seems to focus on text data, with source and target strings that could be used for various NLP tasks such as text generation or machine translation.