Now, at least in a well disciplined human, we can catch when our gut feeling was wrong when the 'create a reasonable argument' process fails. I guess I wonder h

Now, at least in a well disciplined human, we can catch when our gut feeling was wrong when the 'create a reasonable argument' process fails. I guess I wonder how well a LLM can catch that and correct it's thinking. Now I've seen in some models where it figures out it's wrong, but then gets stuck in a loop. I've not really used the larger reasoning models much to see their behaviors.