Fetching the paper…

ReaL: Efficient RLHF Training of Large Language Models with Parameter Reallocation · Around