AI Glossary · Last reviewed August 2026

RLHF

· Reinforcement Learning from Human Feedback

Hand-written by a real person. Reviewed against current practice in August 2026.

Definition

A training method where humans rank model outputs to teach the model what good looks like.

Why it matters

RLHF is the training technique that made ChatGPT feel helpful instead of robotic. It uses human feedback to teach models what good responses look like - not just grammatically correct ones, but actually useful, safe, and aligned ones.

Understanding RLHF helps you appreciate why different models feel different to use, and why some handle sensitive topics more carefully than others.

How it works

4 steps

STEP 01

Model generates responses

The AI model produces multiple candidate answers to the same prompt.

STEP 02

Humans rank the outputs

Human reviewers compare the responses and rank them from best to worst based on helpfulness, accuracy, and safety.

STEP 03

Reward model learns preferences

A separate reward model is trained on these human rankings to predict which responses humans would prefer.

STEP 04

Policy is optimized

The main model is fine-tuned using reinforcement learning to maximize the reward model's score - producing more human-preferred responses.

Related terms

From the glossary

Fine-tuning

LLM

Frequently asked questions

What kind of feedback does RLHF use?+

Human raters compare pairs of model outputs and select the better one. This preference signal trains a reward model, which then guides reinforcement learning to make the base model produce outputs more like the preferred ones.

Is RLHF used in all major models?+

Most frontier chat models, including GPT-4, Claude, and Gemini, use some form of human feedback alignment. The exact method varies and newer techniques like DPO and RLAIF are also emerging.

What are the limitations of RLHF?+

It is expensive to collect human preferences at scale, annotator disagreements introduce noise, and models can learn to game the reward model rather than genuinely improving.

New to RLHF?

See the tools that use it.

The fastest way to understand RLHF is to see it inside real products. Browse hand-reviewed tools that put it to work, each one checked by a person before it was listed.

Browse hand-reviewed AI tools