hanane/attributionBench_augmentedselectErrortypesbalanced

该数据集包含问题、断言、原始断言字符串、响应、参考文献、引用链接、网页参考文献、归因标签、源数据集、ID、NLI分数、对齐分数、示例类型、原始标签、标签切换、错误类型、选中内容及其角色、拒绝内容及其角色、消息及其角色、选中得分和拒绝得分等字段。数据集分为训练集和验证集，训练集包含30212个示例，大小为236904538字节；验证集包含7552个示例，大小为58077928字节。总下载大小为150942140字节，数据集总大小为294982466字节。

The dataset includes fields such as question, claim, raw claim string, response, references, citation links, webpage references, attribution label, source dataset, ID, NLI score, alignment score, example type, original label, label switch, error type, chosen content and its role, rejected content and its role, messages and their roles, score chosen, and score rejected. The dataset is split into training and validation sets, with the training set containing 30,212 examples and being 236,904,538 bytes in size; the validation set contains 7,552 examples and is 58,077,928 bytes in size. The total download size is 150,942,140 bytes, and the total dataset size is 294,982,466 bytes.