Applying knowledge or rules learned from one model to improve performance in a different, often smaller model.