AIF-C01 Question Bank · Latest 50 Questions Free Preview (Page 4/5)
Questions 31–40 of the latest 50 AIF-C01 questions (5 pages total), with answers and explanations, no login required. Practice all 435 questions online.
Question 31 · No. 416 · Single choice
An AI practitioner is developing a recommendation system. The AI practitioner wants to document a business problem, data assumptions, training considerations, and usage risks. The company must follow guidelines for transparency and governance. Which Amazon SageMaker AI feature will meet these requirements?
- A. Model Registry
- B. Model Cards
- C. Model Monitor
- D. Model Dashboard
Answer: B
Explanation: Amazon SageMaker Model Cards are specifically designed to document and track information about machine learning models, including business problems, data assumptions, training considerations, and intended uses and risks. They provide a standardized way to document model information for transparency and governance purposes, which directly matches the requirements in the question.
Question 32 · No. 417 · Single choice
An AI practitioner wants to use Retrieval Augmented Generation (RAG) architecture. However, company leadership thinks RAG will make the architecture too complex. Which response will support the AI practitioner's recommendation to use RAG?
- A. RAG improves large language model (LLM) inference time.
- B. RAG improves model training speed.
- C. RAG decreases hallucinations.
- D. RAG decreases development controls.
Answer: C
Explanation: RAG (Retrieval Augmented Generation) is known to decrease hallucinations in LLM outputs by grounding responses in retrieved, factual information from a knowledge base. This addresses a key concern with LLMs generating inaccurate or fabricated information, making it a strong supporting argument for adopting RAG architecture.
Question 33 · No. 418 · Single choice
A company plans to use a dataset to train a new AI model. The company needs to determine whether a bias against a specific user group exists in the dataset before the company trains the model. Which solution will meet these requirement?
- A. Use Amazon SageMaker Clarify to compute pre-training bias metrics.
- B. Use Amazon SageMaker Model Cards to compute pre-training bias metrics.
- C. Use Amazon SageMaker Clarify to compute post-training bias metrics.
- D. Use Amazon SageMaker Model Cards to compute post-training bias metrics.
Answer: A
Explanation: Amazon SageMaker Clarify is the appropriate service for detecting bias in datasets. The company needs to check for bias before training the model, which means they need pre-training bias metrics. SageMaker Clarify can compute both pre-training and post-training bias metrics, while SageMaker Model Cards is used for documenting model information and does not compute bias metrics. Therefore, option A (Use Amazon SageMaker Clarify to compute pre-training bias metrics) is the correct answer.
Question 34 · No. 419 · Single choice
Which benefit does a company gain by using ongoing pre-training to fine-tune a foundation model (FM)?
- A. Decreased model complexity
- B. Improved model performance over time
- C. Decreased time required to train the model
- D. Reduced model training cost
Answer: B
Explanation: Ongoing pre-training involves continuously updating a foundation model with new data over time. This process allows the model to learn from the latest information, adapt to changing patterns, and refine its understanding of various tasks. As a result, the model's performance gradually improves over time, making it more accurate and effective. Option A is incorrect because pre-training doesn't necessarily reduce model complexity. Option C is incorrect because ongoing pre-training actually requires continuous training time. Option D is incorrect because ongoing pre-training typically increases costs due to the continuous use of computational resources.
Question 35 · No. 420 · Multiple choice
A company is preparing data to fine-tune a large language model (LLM). The company must follow the principles of responsible AI. Which actions will meet this requirement? (Choose two.)
- A. Include demographically diverse data sources to improve model fairness.
- B. Apply consistent labeling standards across the training dataset.
- C. Focus exclusively on high-volume data regardless of quality.
- D. Use only synthetic data to train the model.
- E. Use fewer labels to reduce ambiguity in data during model training.
Answer: A, B
Explanation: The two correct answers are A and B because they align with the principles of responsible AI. Option A ('Include demographically diverse data sources to improve model fairness') supports fairness and inclusivity, which are core responsible AI principles that help reduce bias in the model. Option B ('Apply consistent labeling standards across the training dataset') ensures data quality and reliability, which is essential for building a trustworthy and accurate model. Options C and D violate responsible AI principles because they either ignore data quality (C) or rely solely on synthetic data which may not represent real-world diversity and could introduce other biases (D). Option E is incorrect because using fewer labels would likely increase ambiguity rather than reduce it, and does not support responsible AI practices.
Question 36 · No. 421 · Single choice
A company has created an AI assistant to provide relevant information to employees about internal processes. The AI assistant uses a fine-tuned foundation model (FM). The company wants to use feedback from employees to further refine the model's responses. Which solution will meet these requirements?
- A. Implement model distillation.
- B. Perform continued pre-training and incorporate feedback from employee ratings.
- C. Use reinforcement learning from human feedback (RLHF).
- D. Enable intelligent prompt routing.
Answer: C
Explanation: The correct answer is C - Use reinforcement learning from human feedback (RLHF). RLHF is specifically designed to refine a model's responses based on human feedback. In this scenario, employees can rate the AI assistant's responses, and that feedback is used to train a reward model that further improves the quality of the model's outputs. Option A (model distillation) is for transferring knowledge between models, not for incorporating feedback. Option B (continued pre-training) is for adapting a model to new domains or data, not specifically for response refinement. Option D (intelligent prompt routing) is for directing prompts to appropriate models, which doesn't address the feedback loop requirement.
Question 37 · No. 422 · Single choice
A company has fine-tuned an Amazon Bedrock foundation model (FM) to produce short document summaries. The company wants an automated metric that compares each model-generated summary with its human-written reference summary. Which metric will meet these requirements?
- A. F1 score
- B. Recall-Oriented Understudy for Gisting Evaluation (ROUGE)
- C. Perplexity
- D. Fréchet inception distance (FID)
Answer: B
Explanation: ROUGE (Recall-Oriented Understudy for Gisting Evaluation) is specifically designed to evaluate the quality of automatic summarization by comparing model-generated summaries against human-written reference summaries. It measures the overlap of n-grams, word sequences, and word pairs between the generated and reference summaries. F1 score is used for classification tasks, Perplexity measures language model performance but doesn't compare to references, and FID is used for evaluating image generation models, not text summarization.
Question 38 · No. 423 · Single choice
A company is building a chatbot. The chatbot uses a large language model (LLM) and answers customer questions about products. The company wants the chatbot to answer only product questions. The company does not want the chatbot to answer questions about other topics. Which solution will meet these requirements with the LEAST operational overhead?
- A. Set guardrails on the LLM prompt template.
- B. Write custom application logic to identify questions about other topics.
- C. Reduce the information the LLM can access.
- D. Set the temperature parameter value to a lower number.
Answer: A
Explanation: Setting guardrails on the LLM prompt template is the most efficient solution with the least operational overhead. By including specific instructions in the prompt that direct the LLM to only answer product-related questions and politely decline other topics, you can achieve the desired behavior without developing additional code (Option B), which would require ongoing maintenance. Reducing information access (Option C) doesn't directly prevent off-topic answers, and adjusting the temperature parameter (Option D) only affects response randomness, not topic relevance. Prompt-based guardrails are a built-in feature that requires minimal configuration and no custom development.
Question 39 · No. 424 · Single choice
Which strategy will evaluate the performance of a foundation model (FM) in real-world applications?
- A. Conducting A/B testing with users in a controlled environment
- B. Human evaluation by subject matter experts
- C. Measuring the model’s accuracy on a training dataset
- D. Analysis of the model’s internal representations and attention patterns
Answer: A
Explanation: A/B testing with users is the most appropriate strategy for evaluating a foundation model's performance in real-world applications. This method involves deploying the model to actual users and measuring their interactions, preferences, and outcomes, providing direct insight into how the model performs under real-world conditions. Option B (human evaluation by SMEs) assesses quality but doesn't capture real-world user behavior. Option C (accuracy on training data) doesn't reflect real-world performance since models are typically evaluated on unseen data. Option D (analysis of internal representations) is an interpretability technique, not a performance evaluation method for real-world use.
Question 40 · No. 425 · Single choice
A financial company uses an ML model to detect potentially fraudulent transactions. The company needs to ensure that some types of predictions receive review by human analysts before the company acts upon the predictions. Which AWS solution will meet this requirement?
- A. Amazon SageMaker Clarify
- B. Amazon SageMaker Ground Truth
- C. Amazon Augmented AI (Amazon A2I)
- D. Amazon SageMaker Model Monitor
Answer: C
Explanation: The correct answer is C - Amazon Augmented AI (Amazon A2I). This AWS service is specifically designed to incorporate human review into machine learning workflows. Amazon A2I allows organizations to set up human review loops for ML predictions, particularly for low-confidence predictions or sensitive use cases like fraud detection in financial services. In this scenario, the financial company needs human analysts to review certain fraud detection predictions before taking action, which is exactly what Amazon A2I is built for. Option A (SageMaker Clarify) is for bias detection and model explainability, Option B (SageMaker Ground Truth) is for data labeling, and Option D (SageMaker Model Monitor) is for monitoring model performance in production.