JaFIn

JaFIn是一个专为日本金融领域大型语言模型（LLM）设计的指令数据集。该数据集由北海道大学信息科学与技术学院和东京大学工程学院的研究人员手工构建，基于多个数据源，包括日本政府网站，提供了丰富的金融知识。JaFIn包含1490个样本，涵盖了从预算和结算到税收系统等多个金融主题。数据集的创建旨在通过指令调优提高LLM在金融领域的适应性和性能，特别是在处理信息提取、摘要和分类等金融实践中的自然语言处理任务。

JaFIn is an instruction dataset specifically tailored for large language models (LLMs) in the Japanese financial domain. This dataset was manually constructed by researchers from the School of Information Science and Technology, Hokkaido University and the School of Engineering, The University of Tokyo. It is built upon multiple data sources including Japanese government websites, and encompasses abundant financial knowledge. JaFIn contains 1,490 samples covering a diverse range of financial topics spanning from budgeting and settlement to tax systems. The dataset was developed to enhance the adaptability and performance of LLMs in the financial domain via instruction tuning, particularly for natural language processing tasks in real-world financial practices such as information extraction, summarization and classification.