# What is Fine-tuning?

Source: https://mushbloom.uk/resources/glossary/what-is-fine-tuning.html

By Mushbloom. Published and updated: 2026-10-12.

## Definition & overview

Fine-tuning is additional training that adjusts a model using task-specific examples. It can improve recurring behaviours or formats; it is not the same as supplying current documents at answer time.

## Analogy & mental model

It teaches the assistant a working style, while RAG gives the assistant the latest handbook.

Use clear baseline tests to decide whether fine-tuning is needed instead of better instructions or retrieval.

## Why it matters

Some repeated behaviours are difficult to achieve consistently with instructions alone. Fine-tuning can help with those patterns, but adding frequently changing facts to training is not a substitute for retrieval. It should follow a measured baseline and a clear reason to change the model.

## Practical business example

A model is trained to classify an established set of ticket types consistently, while current account facts continue to come from authorised retrieval.

Illustrative scenario, not a client case study or promised result.

## How it works

1. Collect high-quality task examples with permitted data and desired responses.

2. Train using a supported method and hold out representative evaluation cases.

3. Compare the resulting behaviour with the baseline before deploying it.

## Core components

- Training examples

- Model adjustment

- Held-out evaluation

Do not use confidential training examples without appropriate rights, controls and a clear understanding of provider data handling.

## Fine-tuning vs RAG

Fine-tuning changes model parameters. RAG supplies retrieved evidence during an interaction and can update facts without repeating the training process.

## Frequently asked questions

What does Fine-tuning mean?
Fine-tuning is additional training that adjusts a model using task-specific examples. It can improve recurring behaviours or formats; it is not the same as supplying current documents at answer time.

How is it different from RAG?
Fine-tuning changes model parameters. RAG supplies retrieved evidence during an interaction and can update facts without repeating the training process.

What is a sensible first step for a small business?
Use clear baseline tests to decide whether fine-tuning is needed instead of better instructions or retrieval. Collect high-quality task examples with permitted data and desired responses.

What limitations should I keep in mind?
Do not use confidential training examples without appropriate rights, controls and a clear understanding of provider data handling.

## Sources

OpenAI: Supervised fine-tuning: https://developers.openai.com/api/docs/guides/supervised-fine-tuning

No rankings, AI citations or commercial outcomes are guaranteed.