15.4 Fine-Tuning

Fine-tuning takes a model that was already pre-trained on huge general text and continues training it on a smaller, focused dataset. This adapts the model to a specific task, domain or style — for example teaching a general LLM to answer customer-support questions for a telecom company in the company's own tone.

Imagine a graduate who knows general medicine, then does a residency to specialise in cardiology. The base knowledge stays, but new focused practice sharpens specific skills. Modern methods like LoRA make this cheaper by updating only a small set of extra parameters instead of the whole model.