AI Practitioner · 28% of the exam

Applications of Foundation Models: free practice questions

5 sample questions from our 14-question bank for this domain — answers and explanations included. These are the same scenario-based style as the real AWS exam.

1. A prompt asks the model to 'think step by step' and show its reasoning before giving a final answer to a complex problem. Which technique is being used?

  • A. Chain-of-thought prompting✓ Correct
  • B. Zero-shot classification
  • C. Quantization
  • D. Data labeling
Explanation

Encouraging the model to reason through intermediate steps before answering is chain-of-thought prompting, which often improves performance on complex reasoning tasks. The other options are unrelated: zero-shot classification (B), model compression (C), and dataset preparation (D).

2. Which parameter limits the LENGTH of a model's generated response?

  • A. Maximum tokens (max output length)✓ Correct
  • B. Temperature
  • C. Top-p only
  • D. The IAM policy attached to the caller
Explanation

The maximum-tokens parameter caps how many tokens the model may generate, controlling response length. Temperature (B) and top-p (C) shape randomness/diversity, not length. An IAM policy (D) governs permissions, not output length.

3. Lowering the TEMPERATURE parameter toward 0 when generating a compliance summary has what effect?

  • A. More deterministic, focused, and repeatable outputs✓ Correct
  • B. More random and creative outputs
  • C. Faster training of the model
  • D. Automatic fine-tuning of the model
Explanation

A low temperature reduces randomness, producing more deterministic, focused, and consistent outputs — desirable for compliance summaries. Higher temperature increases creativity/variety (B). Temperature affects inference, not training (C) or fine-tuning (D).

4. A company wants its generative AI assistant to answer questions using its own up-to-date internal documents WITHOUT retraining the model. Which approach should it use?

  • A. Retrieval-Augmented Generation (RAG), retrieving relevant documents and adding them to the prompt at query time✓ Correct
  • B. Continued pre-training on the entire internet
  • C. Lowering the temperature to zero
  • D. Increasing the number of GPUs only
Explanation

RAG grounds responses by retrieving relevant internal documents and injecting them into the prompt at inference time — no model retraining needed, and easy to keep current. Continued pre-training (B) is heavyweight and unnecessary. Temperature (C) and GPU count (D) don't give the model access to internal knowledge.

5. What role does a VECTOR DATABASE play in a Retrieval-Augmented Generation solution?

  • A. It stores embeddings so the system can find semantically similar content to retrieve as context✓ Correct
  • B. It stores the model's trained weights
  • C. It replaces the foundation model entirely
  • D. It encrypts network traffic
Explanation

A vector database stores embeddings and enables similarity search, so a RAG system can retrieve the most relevant chunks to ground the model. It doesn't store model weights (B), replace the model (C), or handle network encryption (D).

9 more questions in this domain

Practice the full bank with instant grading, flashcards, and a timed mock exam.

Start practicing free