In Sections 3.3 and 3.4, the authors mention linear models for the mapping between the text embeddings and the fMRI. I looked at their repository, but it does not have any code yet. Does anyone have a better idea about these linear models the authors talk about?