Early-exit networks offer accelerated inference and reduced forgetting, enhancing performance in continual learning scenarios.