madrylab/gsm8k-platinum

GSM8K-Platinum是一个经过修订的GSM8K测试集的全版本，GSM8K是一个包含小学生数学文字问题的数据集，旨在提供对数学推理能力的更准确的评估。修订过程中，使用了多种前沿模型对每个单独的示例进行测试，并对至少一个模型出现错误的示例进行了人工检查。修订了标记错误的示例的标签，并删除了确定编写不当的任何问题（通常是因为问题陈述的模糊性）。

GSM8K-Platinum is a revised version of the full test set of GSM8K, a dataset of grade school math word problems, providing a more accurate assessment of mathematical reasoning capabilities. The revision process involved running a variety of frontier models on each individual example and manually reviewing any example for which at least one model made an error. The labels of mislabeled examples were revised, and any question determined to be poorly written was removed, often due to ambiguity in the problem statement.