aiqwe/FinShibainu

FinShibainu数据集是一个专注于韩语金融领域的数据集，主要用于问答、多选和文本生成任务。数据集包含三个配置：mcqa、qa和validation。mcqa配置用于多选问答任务，包含参考文档、问题、选项、推理过程等列；qa配置用于问答任务，包含参考文档、问题、两个答案、偏好等列；validation配置用于验证任务，包含问题、选项、推理过程等列。数据集还提供了相关的引用和引用链接。

The FinShibainu dataset is a large language model training and validation dataset specifically designed for the financial domain, primarily used for question-answering (QA) and multiple-choice question-answering (MCQA) tasks. The dataset includes multiple configurations, each with different features and purposes. The MCQA configuration is used for multiple-choice question-answering tasks, containing features such as reference documents, questions, options, reasoning process, number of options, answers, steps, value assessment, and data type. The QA configuration is used for question-answering tasks, containing features such as reference documents, questions, two different answers (based on and not based on reference documents), preference selection, preference description, value assessment, and data type. The validation configuration is used for model checkpoint testing, containing features such as questions, options, reasoning process, answers, tasks, number of options, and data type. The dataset sources include multiple financial and economic-related reference documents, such as economic and financial terms from the Bank of Korea, regulations and non-periodic publications from KRX, etc.