Unlocking Reasoning Potential in Large Langauge Models by Scaling Code-form Planning | Read Paper on Bytez