What Makes Linguistic Representations Good Models of High-Level Visual Perception in the Human Brain?
This question underlies the study presented in our new preprint. We trained brain-encoding models to predict fMRI responses to naturalistic images from LLM-derived embeddings of image captions.
