I realize this question is extremely petty, but it has professional implications for me. Chat GPT-4 gave my dissertation a B+ and GPT-5 gave it an A. The dissertation is in analytic metaphysics. Which model is more reliable for this kind of cognitive task?