AI engineering glossary · LLM training alignment
What is SFT (supervised fine-tuning)?
Supervised fine-tuning (also called SFT): Continuing training of a pretrained model on labelled input-output examples written or curated by people, using the same next-token loss but only on the desired responses. It teaches format, tone and task behaviour and is normally the first post-training step before preference-based alignment.
Supervised fine-tuning: an example
Fine-tuning on 10,000 customer-support chats where the target is the agent's reply teaches a model to answer in a company's style.
Related terms
More LLM training alignment terms
- Sycophancy
- Synthetic data
- Task arithmetic
- Teacher-student learning
- Tensor parallelism
- Test-time compute scaling
- Trust region policy optimisation
- UltraFeedback
Where you learn it
Supervised fine-tuning is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.