Comment by slama

The title here doesn't seem to match. The paper is called "TopoNets: High Performing Vision and Language Models with Brain-Like Topography"

Even with their new method, models with topography seem to perform worse than models without.