maximedb/twentle_dpo

该数据集包含了三个主要特征：选中的(chosen)、被拒绝的(rejected)和策略相关的(onpolicy)。选中的和被拒绝的特征都进一步包含了内容(content)、角色(role)和思考(thinking)三个子特征，所有这些特征都是以字符串的形式存储。策略相关特征是一个布尔类型。此外，数据集还有一个字符串类型的word特征。整个数据集分为训练集(train)，共有349个示例，数据集大小为16802689字节，下载大小为3042552字节。

The dataset includes three main features: chosen, rejected, and onpolicy. Both chosen and rejected features further consist of sub-features: content, role, and thinking, all stored as strings. The onpolicy feature is a boolean type. Additionally, there is a string type feature called word. The dataset is split into a training set (train) with a total of 349 examples, with a dataset size of 16802689 bytes and a download size of 3042552 bytes.