NVIDIA · NCA-GENL

NVIDIA Certified Associate: Generative AI LLMs practice questions

Newcomers building and using large language models

Pass markNot published
Questions in the exam50–60
Time allowed60 min
Exam fee$125
Practice questions here25

What the NCA-GENL exam covers

Current exam guide
Core machine learning and AI knowledge30%
Software development24%
Experimentation22%
Data analysis and visualisation14%
Trustworthy AI10%

Sample NCA-GENL questions

Five of the 25 practice questions, each with its answer and explanation.

Core machine learning and AI knowledge

What mechanism lets a transformer weigh how much each token relates to every other token?

  1. Max pooling
  2. Self-attention
  3. Gradient clipping
  4. Dropout
Show the answer

B. Self-attention

Self-attention computes relationships between all tokens in the sequence.

Core machine learning and AI knowledge

Why is cross-validation used?

  1. To speed up inference
  2. To encrypt the data
  3. To reduce the vocabulary
  4. To estimate how well a model generalises by testing on several held-out splits
Show the answer

D. To estimate how well a model generalises by testing on several held-out splits

Rotating which data is held out gives a more reliable estimate of performance.

Core machine learning and AI knowledge

What does a text embedding represent?

  1. The meaning of text as a vector of numbers
  2. The author's name
  3. The file size
  4. The text's font
Show the answer

A. The meaning of text as a vector of numbers

Similar meanings produce nearby vectors, which powers semantic search.

Core machine learning and AI knowledge

What problem does retrieval-augmented generation (RAG) mainly address?

  1. Making models smaller
  2. Training models faster
  3. Removing the need for prompts
  4. Answering from current or private information the model was not trained on
Show the answer

D. Answering from current or private information the model was not trained on

RAG supplies relevant documents at query time to ground the answer.

Core machine learning and AI knowledge

What happens during tokenisation?

  1. Images are resized
  2. Data is encrypted
  3. Text is split into units the model processes
  4. The model is compressed
Show the answer

C. Text is split into units the model processes

Tokenisers map text to token IDs.