Fine-tuning
Continuing to train an existing model on your own examples so it adapts to a specific style, format, or task, rather than learning everything from scratch.
Fine-tuning takes a model that has already been trained and continues training it on a smaller set of your own examples, so it adapts to a particular task, tone, or output format. You are nudging an existing model, not building one from nothing.
It is worth being clear about what fine-tuning does and does not do. It is good at teaching a model a style, a structure, or a behaviour, such as always replying in a specific JSON shape or adopting a house voice. It is a poor and expensive way to teach a model new facts, which is usually the job of RAG instead.
Full fine-tuning updates every weight in the model and needs serious hardware. In practice most people use a lighter method such as LoRA, which trains a small set of extra weights and brings fine-tuning within reach of a single consumer GPU.