I am curious if there is a coherent approach to modeling attention for long- or medium-term goal following. Currently, LLMs resemble a fish or a small child in

I am curious if there is a coherent approach to modeling attention for long- or medium-term goal following. Currently, LLMs resemble a fish or a small child in their focus and behavior. Robots with such limitations would be strange to work with.