Supervised Fine-Tuning

Also known as: SFT

Supervised fine-tuning (SFT) updates a model on labeled input–output pairs — often instruction/response chats. It teaches format, helpfulness style, and task behavior more than it dumps large factual corpora into the weights.

A common recipe for domain specialists: CPT on raw text → light SFT on chat so the model keeps domain knowledge while answering cleanly.

← Back to AI Glossary