A reward is the numerical feedback an environment provides after an agent takes an action in a state. In reinforcement learning, rewards define what the agent is encouraged to maximize over time.
In reinforcement learning, the numerical result of taking an action in a state, as defined by the environment.