Not All Rollouts are Useful: Down-Sampling Rollouts in LLM Reinforcement Learning

Open in new window