InFoBench
收藏资源简介:
InFoBench是一个评估基准数据集,包含500条指令和2250条分解后的需求。数据集中的每个样本都包含指令字符串、输入字符串(可选)、分解后的问题列表以及每个分解问题的标签列表。数据字段包括id、子集(Hard_Set或Easy_Set)、类别信息、指令、输入(上下文信息,可以为空字符串)、分解后的问题列表以及对应于分解问题的标签列表。该数据集遵循MIT许可协议。
InFoBench is an evaluation benchmark dataset consisting of 500 instructions and 2250 decomposed requirements. Each sample in the dataset contains an instruction string, an optional input string, a list of decomposed questions, and a list of labels corresponding to each decomposed question. The data fields include id, subset (either Hard_Set or Easy_Set), category information, instruction, input (contextual information which can be an empty string), a list of decomposed questions, and a list of labels corresponding to the decomposed questions. This dataset is licensed under the MIT License.




