Back
advanced
Fine-Tuning Techniques

Fine-Tuning Methods: SFT, PEFT, RLHF, and DPO

Compare the major model-customization methods and when each one is appropriate

26 min read· Fine-Tuning· PEFT· Adapters· Prompt Tuning

Fine-Tuning Methods: SFT, PEFT, RLHF, and DPO

Fine-tuning changes model behavior. It should be used only after prompting, tools, and RAG are not enough.

Method map

MethodBest for
SFTteaching examples of ideal behavior
LoRA/PEFTlow-cost specialization
QLoRAfine-tuning larger models with less memory
RLHFlearning from human preferences
DPOsimpler preference optimization
distillationmoving behavior into a smaller model

Choose by goal

GoalMethod
consistent formatSFT or LoRA
cheaper repeated taskdistillation or LoRA
better preference behaviorDPO/RLHF
new factual knowledgeusually RAG
tool-call styleSFT/LoRA with tool traces

Evals first

Before training:

  • collect representative examples
  • define success criteria
  • run baseline
  • estimate deployment cost
  • plan rollback

Knowledge check

Q1: What is PEFT?

Parameter-efficient fine-tuning, where only a small set of parameters is trained.

Q2: Which method should handle changing facts?

Usually RAG, not fine-tuning.