Math Training Set

该数据集由来自MetaMathQA的10,000个随机样本组成，这些样本是从GSM8K和MATH训练集中增强而来的。此外，该数据集在基于精确答案匹配的MATH-500测试集上进行评估。其任务是数学推理。This dataset comprises 10,000 randomly sampled instances from MetaMathQA, which are augmented based on the GSM8K and MATH training datasets. Additionally, this dataset is evaluated on the MATH-500 test set, where exact answer matching is adopted as the evaluation metric. The core task of this dataset is mathematical reasoning.