Determining the correct sequence of events or screenshots in time, used here to test if models understand action consequences.