prestonfu/e3-math-medhard-Qwen3-1.7B-2k

该数据集包含了多个字段，如解决方案、奖励、长度、正确长度、错误长度、数据来源、提示内容（包括内容和角色）、能力、奖励模型（包括真实情况和风格）、额外信息（包括索引和分割）以及模型答案和在线策略答案（包括答案、计数和等级）。此外，数据集分为训练集的不同难度级别，包括训练集、训练零、训练困难、训练中等和训练简单。每个split都有相应的字节数和示例数。数据集的总下载大小为2.65MB，总数据大小为6.43MB。

The dataset includes multiple fields such as solution, reward, length, correct length, incorrect length, data source, prompt content (including content and role), ability, reward model (including ground truth and style), extra info (including index and split), and model answers and onpolicy answers (including answer, count, and grade). Additionally, the dataset is split into different difficulty levels of the training set, including train, train_zero, train_hard, train_medium, and train_easy. Each split has its corresponding number of bytes and examples. The total download size of the dataset is 2.65MB, and the total data size is 6.43MB.