Concept
Supervised fine-tuning
Supervised fine-tuning (SFT) is the step where you take a pretrained language model and continue training it on curated input-output pairs, using the same next-token cross-entropy loss as pretraining. The difference is the data: instead of raw web text, you feed it examples…
The rest of “Supervised fine-tuning” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→