Commonsense Reasoning

A research direction in AI referring to the ability of models to reason using implicit knowledge about how the world works, involving social commonsense, physical commonsense and related domains; it is a weak point of current large models and vision-language models.