Exponential advantage in continuous-variable quantum state learning

Abstract: We consider the task of learning quantum states in bosonic continuous-variable (CV) systems. We present a concrete experimentally feasible protocol that utilizes entangled measurements and reflected states to enable efficient learning of the characteristic function of CV quantum states with a sample complexity independent of the number of modes $n$. We then prove that any general adaptive learning scheme without entangled measurements requires a sample complexity that is exponential in $n$ to accomplish the same task, demonstrating an exponential advantage provided by entangled measurements. Remarkably, we also prove that any general adaptive entanglement-assisted scheme also requires an exponential sample complexity if access to reflected states is prohibited. Hence, we establish a rigorous exponential advantage from entangled measurements and access to reflected states in learning CV quantum states.