Experiments show that KAN is robust to reward noise, with its advantages remaining stable when the reward signal has varying degrees of noise.