Reinforcement Learning from Human Feedback — a training technique that aligns model outputs with human preferences.
Adhering to complex, structured, or constrained instructions