Arithmetic Tasks Dataset

该数据集是一组专为算术任务设计的多样化数据集，包含了从100万到500万、1000万、2500万以及5000万条记录的不同规模，并附带了一个包含9592个测试案例的评估数据集。该数据集是从与训练数据集相同的分布中生成的，且与之独立，它作为评估MathGLM模型在算术任务上性能的基准。数据集的规模从100万条记录到5000万条记录不等，其任务定位于解决算术问题。

This dataset is a diverse collection specifically designed for arithmetic tasks, with multiple scales including 1 million to 5 million, 10 million, 25 million, and 50 million records, and it is accompanied by an evaluation dataset containing 9592 test cases. The evaluation dataset is generated from and independent of the same distribution as the training dataset, and serves as a benchmark for evaluating the performance of the MathGLM model on arithmetic tasks. The dataset has sizes ranging from 1 million to 50 million records, and its tasks are targeted at solving arithmetic problems.