非变量泛化【英文版】

英文标题：Out-of-Variable Generalization中文摘要：本文旨在研究 AI 智能代理器对于陌生和新的因变量的预测和探索问题，提出了一种基于先前经验的较为有效的处理方法，并能够实现零样本预测。英文摘要：The ability of an agent to perform well in new and unseen environments is acrucial aspect of intelligence. In machine learning, this ability is referredto as strong or out-of-distribution generalization. However, simply consideringdifferences in data distributions is not sufficient to fully capturedifferences in environments. In the present paper, we assay out-of-variablegeneralization