Robust Reinforcement Learning from Corrupted Human Feedback Alexander Bukharin

Reinforcement learning from human feedback (RLHF) provides a principled framework for aligning AI systems with human preference data.