Fetching the paper…

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