NCA-GENL Core Machine Learning and AI Knowledge Practice Question
A researcher is fine-tuning a large language model using PEFT (Parameter-Efficient Fine-Tuning) techniques. Which method is specifically designed to inject trainable low-rank matrices into the transformer layers to reduce the number of trainable parameters?
⚠ Common exam trap
Candidates often confuse LoRA with full fine-tuning or quantization techniques like QLoRA, mistakenly believing that LoRA changes the original weights of the frozen pre-trained model directly during the update process.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
LoRA
Low-Rank Adaptation (LoRA) is the standard PEFT method for injecting trainable matrices into frozen pre-trained models. By minimizing the number of updated parameters, it significantly reduces VRAM requirements and storage overhead during the fine-tuning process. This technique is vital for deploying custom LLMs on resource-constrained hardware, as it maintains model performance while drastically simplifying the memory demands of the training pipeline.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Prefix Tuning
Why it's wrong here
Prefix tuning prepends learnable vectors to the input of each transformer layer. While effective, it modifies the attention mechanism's input space rather than injecting low-rank matrices into the weight layers themselves. This approach can be more computationally expensive and harder to scale than LoRA for large models.
- ✓
LoRA
Why this is correct
LoRA uses low-rank decomposition to represent weight updates as the product of two smaller matrices. This approach drastically reduces the total number of trainable parameters, allowing for efficient fine-tuning on consumer-grade or limited-memory GPUs without sacrificing the quality of the original pre-trained model weights.
- ✗
Prompt Tuning
Why it's wrong here
Prompt tuning involves learning a set of continuous soft prompts that are appended to the input data. Unlike LoRA, it does not involve modifying the internal weight matrices of the transformer blocks, meaning it provides different trade-offs regarding model capacity and training stability compared to weight-injection methods.
- ✗
Full Fine-Tuning
Why it's wrong here
Full fine-tuning updates every parameter in the pre-trained model. This method is highly resource-intensive and requires massive VRAM for storing optimizer states and gradients for all weights. It is the opposite of parameter-efficient methods and is often impractical for large-scale language models on standard enterprise GPU hardware.
About these practice questions
This NCA-GENL question is part of Courseiva's 367-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official NVIDIA exam blueprint
This NCA-GENL practice question is part of Courseiva's free NVIDIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the NCA-GENL exam.