recursal/longbench-v2

该数据集包含多种自然语言处理任务的配置，包括多项选择题、问题回答、文本分类、表格问题回答等。数据集使用英语，内容涉及长文本上下文和推理，数据规模小于1K。数据集遵循Apache-2.0许可。不同的配置对应不同的数据文件，涵盖了学术、金融、政府、法律等多个领域。这个数据集似乎是为了多任务学习而设计，特别关注长文本上下文理解和推理。

The dataset comprises configurations for various natural language processing tasks such as multiple-choice, question-answering, text-classification, and table-question-answering. The data is in English, focusing on long context and reasoning, with a size category of n<1K. It is licensed under Apache-2.0. Different configurations correspond to different data files, covering domains like academic, financial, government, and legal. This dataset appears to be designed for multitask learning with an emphasis on understanding and reasoning in long contexts.