Define supervised fine-tuning (SFT) in one breath.

Attempt it

SFT is continued next-token training on a curated set of (prompt, response) demonstrations, run on top of a pretrained base model. Mechanically it is the same causal-LM cross-entropy loss as pretraining, but two things differ: the data is small and structured (instruction-following demos, chat conversations, tool-call traces) and the loss is typically masked so only response tokens contribute. The goal is not to teach new facts, the base already has them, but to teach the model the FORMAT and behaviour of being a helpful assistant: when to answer, how to structure replies, how to follow the chat template.

Your answer

Quick reference

TL;DR

SFT is continued next-token training on curated (prompt, response) demos, with loss masked to response tokens, to install assistant behaviour on a pretrained base.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Imagine a brilliant student who has read every book in the library but has never been to a job interview. They know everything but they ramble, they speak out of turn, they cite obscure sources when a one-line answer would do. SFT is a weekend with a tutor who shows the student a few hundred good interview answers. The tutor never teaches new facts; the student already knows the material. What changes is style and shape, when to be concise, when to elaborate, when to refuse, when to ask. By Monday the student answers in the right format. Same knowledge, brand new presentation. That polish is exactly what SFT installs on top of a pretrained base.

Key concepts

Supervised fine-tuning Instruction tuning Chat template Prompt loss masking

Quick scan · 2–4 min

Everything important, quickly.

6 min: SFT definition + same loss as pretraining + masked to assistant tokens + low LR vs pretraining + behaviour vs knowledge distinction + dataset-quality vs quantity + where SFT sits in the alignment stack.

Side-by-side

Dimension	Pretraining	SFT
Data shape	Raw text, no labels	Curated chat-format demos
Data volume	Trillions of tokens	Thousands to low millions
Loss	Next-token CE, every token	Next-token CE, assistant only
Learning rate	1e-4 to 6e-4	1e-5 to 2e-5
What changes	World knowledge	Format and behaviour

Real-world examples

Real products, models, and research that use this idea.

Llama 4 Maverick's post-training begins with a large SFT stage on curated demonstrations before any preference optimisation.
DeepSeek V4 publishes an SFT recipe using around 2e-5 learning rate on a multi-million example chat dataset.
The LIMA paper trained a 65B base on just 1,000 SFT examples and produced competitive instruction-following, the classic demonstration that quality beats quantity.
Hugging Face's TRL SFTTrainer is the default tool used across the open ecosystem; it handles chat template rendering and prompt masking automatically.
OpenAI fine-tuning endpoints (and Anthropic's, and Mistral's) all expose what amounts to an SFT API: upload chat-formatted examples, the backend runs masked next-token training at low LR.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QWhy does prompt loss masking matter so much in practice?

Sign in or go Pro to see the approach for this follow-up.

QHow does LIMA's 1,000-example result reframe SFT data thinking?

Sign in or go Pro to see the approach for this follow-up.

QWhen would you choose continued pretraining over SFT for domain adaptation?

Sign in or go Pro to see the approach for this follow-up.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.