Comment by Seattle3503

I was speaking of reinforcement learning with verifiable rewards. It's not clear to me from your comment why re-writing python couldn't be in the set of training tasks used during RLVR.