Question 39
Domain 1: Generative AI with LLMs and PromptingWhat is the main advantage of Parameter-Efficient Fine-tuning (PEFT) methods like LoRA?
Correct answer: B
Explanation
PEFT methods like LoRA update only a small set of parameters instead of the full model, so they "reduce computational requirements" and memory use during training. This lets the model keep strong task performance while avoiding the cost of full fine-tuning.
Why each option is right or wrong
A. Better accuracy than full fine-tuning
B. Reduces computational requirements while maintaining performance
LoRA is designed to inject low-rank trainable matrices into a pretrained model while freezing the original weights, so only a small fraction of parameters are updated during training. In the original LoRA formulation, this cuts the number of trainable parameters by orders of magnitude and lowers GPU memory and optimizer-state overhead, yet empirical results show performance remains comparable to full fine-tuning on the target task.
C. Eliminates need for pre-trained models
D. Faster inference speed