Evolution Strategy

Because it maximizes a specific scalar score (reward function or -loss), it can be viewed as both model optimization and an RL algorithm.