ETH Zurich's new transformer architecture enhances language model efficiency, preserving accuracy while reducing size and computational demands. Skip to main content.