provide SAGE with a math problem to solve. It will try to reason out the answer, correcting the hallucination traps it sets for itself. PS! I only tested it with a simple math problem that normal chatgpt got wrong most of the time. (Cost too much here to test more)