tuhink/hacking-rewards
收藏Hugging Face2024-12-20 更新2024-12-21 收录
下载链接：
https://hf-mirror.com/datasets/tuhink/hacking-rewards
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含多个特征，主要用于训练模型。其中chosen和rejected是两个列表类型的特征，每个列表包含content和role两个子特征，数据类型均为字符串。此外，数据集还包含source、reward_chosen和reward_rejected三个特征，数据类型分别为字符串和浮点数。数据集仅包含一个训练集分割，共有77,016个样本，总大小为416,854,646字节，下载大小为209,764,031字节。
The dataset contains multiple features, primarily used for training models. Among them, chosen and rejected are two list-type features, each containing content and role sub-features, both of which are of string type. Additionally, the dataset includes source, reward_chosen, and reward_rejected features, with data types of string and float64 respectively. The dataset contains only one training split, with a total of 77,016 examples, a total size of 416,854,646 bytes, and a download size of 209,764,031 bytes.
提供机构：
tuhink
