Learning mode's policy is insufficient