# RLAIF

Uses an AI model, rather than humans, to produce the preference labels used for alignment.

## What it is used for

Scaling alignment data when human labelling is too slow or costly.