JGLUE

该数据集名为JGLUE，旨在评估多语种大型语言模型（LLM）在英语自然语言理解任务及其相应日语任务之间的跨语言迁移性能。数据集包含了多种指令模板，以确保对每个任务进行公平的评价。此外，数据集中还包含了用于限制输出格式的正则表达式，并旨在评估模型在零样本和微调设置下的性能。该数据集所针对的任务是自然语言理解（NLU）。

This dataset, named JGLUE, aims to evaluate the cross-lingual transfer performance of multilingual Large Language Models (LLMs) between English natural language understanding tasks and their corresponding Japanese counterparts. It incorporates multiple instruction templates to ensure fair evaluation for each task. In addition, the dataset also includes regular expressions for constraining output formats, and it is designed to assess model performance under both zero-shot and fine-tuning settings. The tasks targeted by this dataset belong to the field of Natural Language Understanding (NLU).