Guji_MATH
收藏资源简介:
Guji_MATH数据集是基于《算经十书》构建的,用于评估经典文本的基准。数据集包含了538个数学问题,这些问题从8部经典文本中提取出来,形成了一个以“问题-答案-解决方案”框架为中心的结构化数据集,并补充了问题类型和难度级别。该数据集旨在评估主流推理模型在古典中文的独特语言限制下的数学问题解决能力。数据集采用机器辅助注释和人工验证的方式,对数学问题进行提取和标注,为从古代文本中挖掘数学知识并传播传统文化提供了方法论支持,同时为评估推理模型的跨语言和跨文化能力提供了新的视角。
The Guji_MATH dataset is constructed based on *Ten Books on Arithmetic (Suanjing Shishu)*, a collection of ancient Chinese mathematical classics. It contains 538 mathematical problems extracted from 8 classical texts, forming a structured dataset centered on the "problem-answer-solution" framework, with supplementary annotations including problem types and difficulty levels. This dataset aims to evaluate the mathematical problem-solving capabilities of mainstream reasoning models under the unique linguistic constraints of classical Chinese. The extraction and annotation of its mathematical problems were completed through machine-assisted annotation and manual verification. It provides methodological support for excavating mathematical knowledge from ancient texts and disseminating traditional Chinese culture, while offering a novel perspective for evaluating the cross-linguistic and cross-cultural capabilities of reasoning models.




