Skip to content

Model Training & Adaptation

Reinforcement learning from human feedback

RLHF

Training that uses human preference judgments to shape model behavior.

Example

People rank two answers, and those preferences help train the assistant's behavior.

Why people use it

Knowing how “Reinforcement learning from human feedback” works helps teams adapt models deliberately and diagnose training problems.

What you'll hear

“Would Reinforcement learning from human feedback improve the model for our specific use case?”

Related terms