What is Model Fine-Tuning

Definition

Model fine-tuning adapts a pre-trained AI model to a specific task or domain by continuing its training on a curated, task-specific dataset, adjusting the model's weights so it produces more accurate, consistent, and specialised outputs than the base model alone.

« Back to Glossary Index

Specialises a base model to a domain, task, or style
Improves accuracy and consistency beyond prompting alone
Bakes desired behaviour into the model, shortening prompts
Builds on a pre-trained base, needing far less data than from scratch

Real World Example

A legal-tech company fine-tunes a base model on its annotated contract clauses so the model reliably extracts and classifies clause types in the firm's exact format, which prompting alone could not do consistently.

FAQs

When should you fine-tune a model?

When you need consistent specialised behaviour and have quality task data, and prompting cannot reach the required accuracy or style.

How does fine-tuning differ from RAG?

Fine-tuning changes the model's weights to embed behaviour, while RAG supplies external knowledge at query time without retraining.

What data is needed for fine-tuning?

A curated, representative, high-quality dataset of examples for the target task or domain.