Fetching the paper…

Tapered Off-Policy REINFORCE: Stable and efficient reinforcement learning for LLMs · Around