LocalResearchGroup/split-NuminaMath-CoT

该数据集包含四种配置：100k、10k、1k和full，每种配置都有训练集和测试集。数据集的特征包括信息源（source）、问题（problem）、解决方案（solution）、提示（prompt）和响应（response）。其中，100k配置的训练集包含90000个示例，测试集包含10000个示例；10k配置的训练集包含9000个示例，测试集包含1000个示例；1k配置的训练集包含900个示例，测试集包含100个示例；full配置的训练集包含773544个示例，测试集包含85950个示例。

The dataset consists of four configurations: 100k, 10k, 1k, and full, each with a training set and a test set. The features of the dataset include source, problem, solution, prompt, and response. The 100k configuration has a training set with 90000 examples and a test set with 10000 examples; the 10k configuration has a training set with 9000 examples and a test set with 1000 examples; the 1k configuration has a training set with 900 examples and a test set with 100 examples; the full configuration has a training set with 773544 examples and a test set with 85950 examples.