In my experience, lean will show that it's correct, but does it not lose the mathematical intuition that led to the result? As far as my experience goes, that's really hard to encode in lean itself.
Could we maybe get more information about the problem from the LLM trace itself here?
Could we maybe get more information about the problem from the LLM trace itself here?