What is fine-tuning? Extra training that shapes an already-trained model for a specific job or style

Extra training that shapes an already-trained model for a specific job or style.

Fine-tuning takes a model that has finished pretraining and trains it a bit more on a smaller set of examples, so it fits a particular task or tone. It doesn't start over; it adjusts what the model already knows. Teaching a model to answer questions helpfully, using model answers, is one kind of fine-tuning.

Often the original numbers stay frozen and only a small add-on set is trained. LoRA, published in 2021, works this way, which makes fine-tuning even big models fairly cheap. Fine-tuning is good at changing style and format; for adding new facts, RAG, which looks things up and answers with the source open, is often a better fit.

Think of it like this

Say a friend who's already great at soccer is asked to play goalkeeper. They don't relearn soccer from scratch; they add a few weeks of goalkeeper practice. Fine-tuning likewise adds one skill on top of what's already there.

For example

From the same base model, fine-tuning can produce a coding helper, a math tutor and a friendly chat buddy as separate versions.

Common mix-ups

Mix-upFine-tuning means training the model again from scratch.
ActuallyLoRA, from 2021, cut the numbers that need training when adjusting a GPT-3-sized model to about one ten-thousandth of the original. It didn't slow down answering either — a completely different scale from teaching it all over again.
Mix-upIf you want to add new facts, fine-tuning is the best way.
ActuallyIf a bus timetable changes, re-tuning the model every time is a hassle; swapping in the new timetable file is much quicker and easier to check. Facts added by fine-tuning are also hard to trace back to a source.

Try it in class

For teachers at the start of a lesson — or for students working on their own.

Question: For a "school lunch menu helper", would you pick fine-tuning or RAG?
Activity: Each group picks an AI helper to build (a maths tutor, a friendly guide, today's lunch menu…) and writes whether what they want to change is tone and format or new facts. Mark tone as fine-tuning and new facts as RAG, then explain why. Open the FLAN record (E0131) to see when "tuning a model to follow instructions" appeared.