In other words… more dimensions for a bigger set of vectors?
Not just more dimensions, but off-the-shelf multimodal encoders