I’ve had the same idea to improve reasoning of LLM. Glad you made it.