qwselfcorr/numia_math_and_augmath_cor20k_20k_20k_2e6_4ep_tmp07_turn2

该数据集包含以下字段：索引（idx）、提示（prompt）、答案序列（answers）、正确答案（gt）、首次奖励（first_rewards）、代理标签（proxy_label）和二次奖励（second_rewards）。数据集分为训练集，大小为53855256字节，共有20000个样本。数据集的下载大小为21146486字节，完整大小为53855256字节。

The dataset includes the following fields: index (idx), prompt (prompt), answer sequence (answers), ground truth (gt), first reward (first_rewards), proxy label (proxy_label), and second reward (second_rewards). The dataset is split into a training set, which is 53855256 bytes in size and contains 20,000 samples. The download size of the dataset is 21146486 bytes, and the full size is 53855256 bytes.