I’m going through an AI course on Udemy right now and they have a similar example. Navigating a maze, with the negative reward too high results in the agent jumping into a fire pit instead of trying to win the game. 😂