Comment by Davidzheng

And we can train models specifically at math proofs? I think only difference is that math is bigger....