# KTO

Aligns a model from simple good or bad labels on single answers, instead of paired comparisons.

## What it is used for

Alignment when only thumbs-up or thumbs-down data is available.