My best guess as a noob is that the vector spaces allow for unbounded contextualization. As long as the training set is large enough, it can 'infer' anything.

My best guess as a noob is that the vector spaces allow for unbounded contextualization. As long as the training set is large enough, it can 'infer' anything. Proofread has a spot in that space, and layers allow patterns like terminology consistency to be expressed so your query will now tap into a subspace that will infer tokens based on whatever consistency patterns were ingested with proofreading texts.