If you do not have that, and large language models do not have that, then you do not know what you do not know. You cannot do belief updating. All you can do is produce outputs for any given input.