McGill-NLP/book_author_qa

该数据集包含多个配置，每个配置针对不同的任务类型和训练场景。具体包括：FLAN少样本学习配置、FLAN零样本学习配置、不同计数水平的忘记问题配置（包括原句和释义）、图卢全测试数据集配置和世界事实数据集配置。每个配置都包含问题和答案字符串、任务名称、任务来源、模板类型、模板索引、响应字符串和答案是否在响应中的布尔值等特征。此外，忘记问题配置还包括书籍、作者和计数等特征。

The dataset consists of multiple configurations, each tailored for different task types and training scenarios. It includes FLAN few-shot learning configuration, FLAN zero-shot learning configuration, forget questions with different count levels (including original and paraphrased sentences), Tulu full test dataset configuration, and World Facts dataset configuration. Each configuration contains features such as question and answer strings, task name, task source, template type, template index, response string, and a boolean indicating whether the answer is in the response. Additionally, the forget questions configuration includes features like book, author, and count.