Our paper, “Flexible task abstractions emerge in linear neural networks with fast and bounded units,” won a spotlight at NeurIPS! Check out the article and accompanying video.