Authors claim their mechanism is more computationally efficient (both asymptotically and on current hardware) than the Transformer attention, while performing m

Authors claim their mechanism is more computationally efficient (both asymptotically and on current hardware) than the Transformer attention, while performing much the same role, yes