Jan_Kulveit comments on Reward is not the optimization target