KMIR

该数据集是一个用于评估语言模型在知识记忆、识别和推理方面的基准测试，涵盖了各种类型的知识和相应的问题答案。特别地，该数据集专注于三元组补全关系类型，用以评估模型对知识的记忆能力。该数据集规模宏大，提供了超过14万条的样本用于训练，其任务是进行知识记忆能力的评估。

This dataset is a benchmark designed to evaluate language models' capabilities in knowledge memorization, recognition and reasoning, covering various types of knowledge and their corresponding question-answer pairs. Specifically, this dataset focuses on the triple completion relation type to assess models' knowledge memorization abilities. With a large scale, this dataset provides over 140,000 training samples, and its core task is to evaluate the knowledge memorization capability of language models.