Asynchronous RLHF: Faster and More Efficient Off-Policy RL for Language Models

Open in new window