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

OffRL
Re-assign community