Parameter-efficient fine-tuning

Fine-tuning techniques that modify only a subset of parameters, reducing computational cost. PEFT techniques like LoRA produce smaller artifacts than full fine-tuning, but ownership and portability of these artifacts still requires contractual clarity.