Fine-Tuning (SFT / LoRA)
Process of updating LLM weights on specialized task datasets to adjust domain behavior or response formats.
Mechanism & Definition
The process of continuing training on a pre-trained foundation model using a task-specific dataset. Supervised Fine-Tuning (SFT) updates model weights directly, whereas Parameter-Efficient Fine-Tuning (PEFT/LoRA) freezes foundation weights and trains low-rank adapter matrices.