Fine-tuning

Continuing training on a pretrained model with a smaller, targeted dataset, so it specializes without relearning language from scratch.

Companion explanation in Step by Token, chapter 14.