qwselfcorr/math_and_augmath_cor20k_20k_2e6_2ep_tmp07_turn2
收藏Hugging Face2025-02-01 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/qwselfcorr/math_and_augmath_cor20k_20k_2e6_2ep_tmp07_turn2
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了一系列的特征字段，如索引、提示、答案序列、真实标签、首次奖励、代理标签和二次奖励。这些字段的名称暗示该数据集可能用于某种序列生成或分类任务，其中可能涉及到奖励机制。数据集被划分为训练集，包含57284515字节，共20000个样本。然而，README文件中并没有提供具体的数据集描述。
The dataset includes a series of feature fields such as index, prompt, answer sequence, ground truth label, first reward, proxy label, and second reward. The names of these fields suggest that the dataset might be used for some kind of sequence generation or classification task, possibly involving a reward mechanism. The dataset is split into a training set, containing 57284515 bytes and a total of 20,000 samples. However, the README file does not provide a specific dataset description.
提供机构：
qwselfcorr
