For those of us not steeped in AI culture, this appears to be short for "Reinforcement learning from human feedback".