Parameter-Efficient Fine-Tuning (LoRA)

LoRA adapts models by training small added matrices instead of all weights.

What is the main benefit of LoRA (Low-Rank Adaptation)?