Chumor

该数据集名为Chumor，来源于类似中国Reddit的平台——糗事百科，包含了带有解释的笑话，旨在帮助评估大型语言模型对幽默理解的能力。Chumor是首个中文幽默解释数据集，为最先进的语言模型提供了一个具有挑战性的基准。该数据集规模涵盖了1951个笑话，总计超过151,730个中文字符的解释内容，其任务是对幽默理解和解释进行评估。

This dataset, named Chumor, is derived from QiuShi BaiKe, a Chinese Reddit-like platform. It collects jokes paired with their corresponding explanations, with the goal of assisting in evaluating large language models' (LLMs') capacity for humor understanding. As the first Chinese humorous explanation dataset, Chumor offers a challenging benchmark for state-of-the-art language models. The dataset encompasses 1,951 jokes, with a total of over 151,730 Chinese characters in their accompanying explanatory content, and its core task focuses on evaluating humor comprehension and explanation.