Supervised Fine-tuning
Also called SFT
Supervised fine-tuning adapts a pretrained model using examples of desired behavior. For language models, training can teach the model to produce target responses or continue supplied text.
[Hugging Face]In practice · hypothetical example
A model is trained on questions paired with carefully written answers in the organization’s preferred style.
[Hugging Face]A little deeper
Prompt-completion and conversational datasets can provide the training examples. The learning objective compares model predictions with the training targets. [Hugging Face]
A common mix-up
SFT chooses actions only from a scalar reward.
Supervised examples provide target outputs to learn from. [Hugging Face]
Which data fits a supervised question-answer fine-tuning task?
Sources & editorial notes
Evidence: supported. Primary-source support for this scoped entry; publication approved by the project owner.
- SFT Trainer ↗ (opens in new tab)Hugging Face · Publication date unknown
Relevant section: Overview; Dataset formats
Last editorial review: 2026-09-13 by project-owner.
First observed in this corpus: Unknown.
Revision history
Revision 2 · Created 2026-09-13 · Updated 2026-09-13
Project owner approved the current content for publication. Existing evidence scope and limitations remain applicable.