croqaz/commonsense-v1

这是一个包含基本常识的数据集，这些知识在过去150年里任何孩子都会知道。它旨在通过提供基本信息来解决大型语言模型（LLMs）缺乏对现实世界理解的问题。数据集有意省略了现代概念（如计算机、互联网、火箭、飞机和汽车）、现代生物学、解剖学和物理学、地理学（因为过去一百年里帝国衰落、新国家建立和城市更名，这些不是永恒的）以及政治等有争议的话题，以保持其永恒和基本的特性。

This is a dataset of basic, timeless knowledge that any child in the last 150 years would know. It aims to address the lack of real-world understanding in large language models (LLMs) by providing fundamental information. The dataset intentionally omits modern concepts (like computers, internet, rockets, planes, and cars), modern biology, anatomy and physics, geography (as empires have fallen, new countries were created, and cities were renamed in the last hundred years, making it not timeless), and controversial topics like politics to maintain its timeless and basic nature.