Full threadvirangjhaveri·Do you reward the RL model based on the token consumption when multiple LLMs complete the task ?View on HN