Techniques
RLHF
Also known as: reinforcement learning from human feedback
In one line
Reinforcement Learning from Human Feedback — the technique that turned raw LLMs into helpful assistants.
What does RLHF mean?
RLHF trains a reward model from human preference rankings, then fine-tunes the LLM to maximise that reward. It's how ChatGPT became useful and polite.
A real-world example
Human raters picked the better of two ChatGPT responses tens of thousands of times to build its reward model.