arg-tech/MysteryZebra

这是一个名为“Mystery Zebra”的数据集，它是作为论文“Lexical Recall or Logical Reasoning: Probing the Limits of Reasoning Abilities in Large Language Models”的一部分而创建的。数据集以.csv格式提供，方便使用。数据集的结构简单明了，易于解析。它包括ID、线索和解决方案网格等字段。ID字段根据测试的不同部分有不同的组成方式，线索则以自然语言的形式给出。该数据集旨在探究大型语言模型的推理能力。

This is the Mystery Zebra dataset created as part of the paper Lexical Recall or Logical Reasoning: Probing the Limits of Reasoning Abilities in Large Language Models. The dataset is provided in a .csv format and consists of ID, clues, and solution grid fields. The composition of the ID field varies depending on the part of the benchmark. Clues are provided in natural language. The dataset is designed to probe the reasoning abilities in large language models.