A random policy is a reinforcement learning policy that selects actions randomly rather than according to learned value or expected reward. It is commonly used for exploration, baselines, or early-stage learning before more informed policies are available.
In reinforcement learning, a policy that chooses an action at random.