Is this using t SNE? Or sth else? I have a feeling similarity is not well defined in whatever space this is using. t SNE is famously unsuited to plot how "close" two points are. it is for clustering