Mechanistic interpretability for steering vision-language-action models Paper • 2509.00328 • Published Aug 30 • 2