MoeReward/combined_preference_dataset_olmoe_base

该数据集包含选定的(chosen)和拒绝的(rejected)两个类别的数据，每个类别下都有内容(content)和角色(role)信息。数据集仅包含训练集split，共有66711个示例。The dataset includes two categories of data: chosen and rejected, each with content and role information. The dataset contains only a training set split, with a total of 66711 examples.