Amazon Web Services Updated AIF-C01 Exam Questions and Answers by akram

Diffusion models are state-of-the-art generative models for creating high-quality, realistic images from textual prompts or other forms of conditioning. These are the foundational technology behind tools like Amazon Bedrock Titan Image Generator and other generative image models.

The verified answer is C. F1 score. This is a classification problem: the model classifies transactions as fraudulent or non-fraudulent. AWS machine learning documentation identifies accuracy, precision, recall, and F1 score as classification evaluation metrics. F1 score is especially appropriate when the evaluation must consider both correct positive predictions and the cost of incorrect positive predictions. AWS defines F1 as the harmonic mean of precision and recall, which means it balances how many positive predictions are correct with how many actual positive cases are found.

The key issue in the question is minimizing false positives. In fraud detection, a false positive means a legitimate transaction is incorrectly classified as fraudulent. AWS documentation explains that precision evaluates how well the model avoids false positives by measuring the percentage of positive predictions that were actually correct. Because F1 includes precision, it penalizes models that generate too many false positives while also considering recall.

Accuracy is not the best answer because it only measures the overall fraction of correct predictions. In fraud detection, the dataset is often imbalanced, so a model can appear accurate while still performing poorly on fraud-related predictions. R-squared and RMSE are regression metrics, not classification metrics. R-squared measures how well a regression model explains variance, and RMSE measures prediction error magnitude for continuous numeric outputs.

Because the company needs a classification metric that reflects correct predictions and penalizes false-positive behavior, F1 score is the strongest option among the choices.

Comprehensive and Detailed Explanation (AWS AI documents):

AWS Responsible AI best practices emphasize that bias should be detected, measured, mitigated, and monitored throughout the ML lifecycle, especially for sensitive domains such as healthcare. When biased outcomes are observed, AWS guidance recommends addressing bias at the data and model level, not only at the output level.

Using Amazon SageMaker Clarify aligns directly with AWS Responsible AI principles because it is designed to:

Detect and quantify bias in datasets and model predictions across sensitive attributes such as demographic groups

Provide pre-training and post-training bias metrics, allowing practitioners to identify where bias originates

Support data-centric mitigation, including improving dataset balance and representativeness

After identifying bias with SageMaker Clarify, collecting additional balanced training data and retraining the model helps ensure that:

The model learns from a more representative dataset

Disparate treatment recommendations based on demographics are reduced

Fairness is improved while maintaining clinical accuracy

Why the other options are not sufficient or aligned with AWS best practices:

B. Prompt engineering can influence outputs but does not address underlying data or model bias and is not sufficient for regulated, high-risk domains like healthcare.

C. Content filtering removes outputs after generation but does not prevent biased decision-making by the model itself.

D. Separate FM endpoints by demographic group increases the risk of reinforcing bias and violates fairness principles rather than mitigating them.

AWS AI Study Guide References:

AWS Responsible AI principles: Fairness and Governance

Amazon SageMaker Clarify: bias detection and mitigation

AWS best practices for ML in high-risk domains such as healthcare

Comprehensive and Detailed Explanation From Exact AWS AI documents:

Prompt engineering is the primary method for controlling tone, style, and behavior of foundation model responses.

AWS generative AI guidance explains that:

Prompts can define tone, voice, and response structure

Iterative refinement ensures consistent outputs

Prompt refinement requires no model retraining

Why the other options are incorrect:

Token limits (A) affect length, not tone.

Batch inferencing (B) affects processing mode, not response style.

Higher temperature (D) increases randomness, reducing consistency.

AWS AI document references:

Prompt Engineering Best Practices

Controlling Model Output Tone

Using Prompts with Foundation Models on AWS