clembench-playpen/DPO_allneg

该数据集包含与游戏相关的信息，具体包括游戏ID、版本、实验信息、回合信息以及模型成功与否的信息等。数据集中的每个样本还包括了提示信息(prompt)、选中的内容(chosen)和拒绝的内容(rejected)。选中的内容和拒绝的内容都包含内容本身、角色和类型三个维度。数据集目前只有一个训练集部分，共有18097个示例。

The dataset includes game-related information such as game ID, version, experiment details, episode information, and whether the model is successful or not. Each sample in the dataset also includes prompt information, chosen content, and rejected content. Both chosen and rejected content consist of three dimensions: the content itself, role, and type. The dataset currently has only one training set part with a total of 18,097 examples.