Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs

• https://arxiv.org/abs/2111.13654

Belief network, Belief system, Language model

This paper proposes methods to detect, update, and visualize the beliefs of language models. This may be used for human-generated texts as well.