AIF-C01 Applications of Foundation Models Practice Test 8 — 15 Questions

Question 1

A team is fine-tuning a foundation model using SageMaker. They want to minimize training time while keeping the model's original knowledge. Which technique is BEST suited?

Accepted Answer

Use Parameter Efficient Fine-Tuning (PEFT) such as LoRA. Parameter Efficient Fine-Tuning (PEFT) methods like LoRA (Low-Rank Adaptation) are best suited because they freeze the pre-trained model weights and inject trainable low-rank matrices into specific layers, drastically reducing the number of trainable parameters. This minimizes training time and computational cost while preserving the model's original knowledge, as only a small fraction of parameters are updated during fine-tuning.

Answer

Use distributed training across multiple GPUs

Answer

Use prompt engineering instead of fine-tuning

Answer

Full fine-tuning on the new dataset

Question 2

A team deployed a text generation model on Amazon Bedrock. They want to monitor for toxic content in model outputs. Which evaluation approach is MOST effective?

Accepted Answer

Use Amazon Bedrock Model Evaluation with toxicity metrics. Amazon Bedrock Model Evaluation with toxicity metrics is the most effective approach because it provides automated, built-in evaluation of model outputs for toxic content using predefined metrics, directly integrated with the Bedrock service. This eliminates the need for manual effort or custom filtering, ensuring consistent and scalable monitoring of harmful content.

Answer

Enable CloudWatch Logs and set a metric filter for toxic words

Answer

Use Amazon SageMaker Ground Truth for human annotation

Answer

Manually review a sample of outputs each week

Question 3

A company is building a multi-modal application that processes images and text to answer questions about product defects. Which foundation model approach is BEST?

Accepted Answer

Use a multi-modal foundation model that processes both images and text. Multi-modal foundation models (e.g., CLIP, Flamingo, GPT-4V) are specifically designed to jointly process and reason over images and text in a unified architecture. This allows the model to directly correlate visual defects with textual descriptions without intermediate lossy transformations, making it the most effective approach for a multi-modal QA task.

Answer

Use an image captioning model and then analyze the caption text

Answer

Use a text-to-image generation model and analyze the generated image

Answer

Use a separate image analysis model and a text model, then combine outputs

Question 4

Which TWO actions can help reduce bias in a foundation model’s outputs? (Choose two.)

Accepted Answer

Fine-tune the model on a balanced, representative dataset. Option A is correct because fine-tuning on a balanced, representative dataset directly addresses bias by exposing the model to fair, diverse examples during training, which adjusts its learned weights and reduces skewed associations in its outputs. Option B is correct because careful prompt engineering with neutral wording avoids injecting biased framing or leading context into the input, so the model is less likely to amplify stereotypes or produce skewed responses. Option C is incorrect because restricting access to a subset of users is an access-control measure that does nothing to change the model's inherent biases. Option D is incorrect because increasing temperature only adds randomness to token sampling, which can make outputs more erratic rather than less biased. Option E is incorrect because a larger foundation model does not inherently reduce bias and may even reproduce or amplify biases present in its larger training corpus.

Answer

Restrict model access to a subset of users

Answer

Increase temperature to add randomness

Answer

Use a larger foundation model

Question 5

Which THREE practices are recommended for responsible AI when deploying foundation models? (Choose three.)

Accepted Answer

Include human review for high-stakes decisions. Option B is correct because responsible AI deployment requires human-in-the-loop oversight for consequential decisions (e.g., hiring, lending, medical triage), so that a qualified person can validate or override model output before it affects someone's rights or safety. Option C is correct because guardrails — input/output filters, content moderation classifiers, and policy enforcement layers — are a standard control to prevent foundation models from generating harmful, unsafe, or policy-violating content. Option D is correct because foundation models can degrade or shift behavior over time due to data drift, concept drift, or upstream model updates, so continuous monitoring of outputs (with metrics, alerts, and retraining triggers) is essential to maintain reliability and fairness. Option A is not recommended: collecting user feedback is a valuable signal for identifying bias and improving models, so avoiding it would reduce accountability rather than enhance responsibility. Option E is not recommended: opaque 'black box' secrecy undermines transparency, explainability, and auditability, which are core responsible AI principles; model internals and documentation should be appropriately disclosed to stakeholders.

Answer

Avoid collecting user feedback to reduce bias

Answer

Use a black box approach to keep model internals secret

Question 6

Refer to the exhibit. This is an Amazon Bedrock invocation request for Claude. What is the purpose of the "stop_sequences" parameter?

Accepted Answer

It tells the model to stop generating when it encounters that sequence. The 'stop_sequences' parameter in Amazon Bedrock's invocation request for Claude tells the model to halt generation as soon as it encounters a specified character sequence. This allows developers to control the output format, such as stopping at a newline or a custom delimiter, ensuring the response ends exactly where intended.

Answer

It specifies a character sequence for the model to include in its response

Answer

It limits the number of tokens in the response

Answer

It controls the randomness of the response

Question 7

Refer to the exhibit. A developer sees this error when calling Amazon Bedrock for inference. What is the MOST likely cause and recommended solution?

Accepted Answer

The request rate exceeds the model's throughput limit; implement retries with exponential backoff. The error indicates a throttling exception from Amazon Bedrock, which occurs when the request rate exceeds the model's throughput limit. The recommended solution is to implement retries with exponential backoff to handle transient rate limits gracefully, as this aligns with AWS best practices for managing API call limits.

Answer

The model ID is incorrect; use a different model

Answer

The prompt is too long; reduce the number of tokens in the prompt

Answer

Increase the max_tokens_to_sample value

Question 8

A company is building a customer support chatbot using Amazon Bedrock. They need to store conversation history for context across sessions. Which AWS service is best suited for this purpose?

Accepted Answer

Amazon DynamoDB. Amazon DynamoDB is the best choice for storing conversation history because it is a fully managed NoSQL key-value and document database that provides single-digit millisecond latency at any scale. It supports flexible schema, which is ideal for storing variable-length chat sessions, and its Time to Live (TTL) feature can automatically expire old conversations to manage storage costs. DynamoDB also integrates natively with AWS Lambda and Amazon Bedrock for real-time retrieval and update of context across sessions.

Answer

Amazon S3

Answer

Amazon RDS

Answer

Amazon ElastiCache

Question 9

A company uses Amazon Bedrock to generate marketing copy. The summaries are too verbose. Which parameter should be decreased to directly limit the length of the output?

Accepted Answer

max_tokens. The `max_tokens` parameter directly controls the maximum number of tokens (words or subwords) in the generated output. By decreasing this value, you explicitly cap the length of the marketing copy, making it less verbose. This is the most direct way to limit output length in Amazon Bedrock and other LLM APIs.

Answer

temperature

Answer

top_p

Answer

frequency_penalty

Question 10

A company uses Amazon Bedrock to generate product descriptions. They need to ensure outputs do not contain offensive language. Which service should they integrate to filter content?

Accepted Answer

Bedrock Guardrails. Amazon Bedrock Guardrails is the correct choice because it is specifically designed to enforce content policies for foundation model outputs, including filtering for offensive language, hate speech, and other harmful content. It integrates directly with Bedrock to apply customizable safety filters and deny topics without requiring additional services or custom code.

Answer

Amazon Comprehend

Answer

Amazon Rekognition

Answer

AWS WAF

Question 11

A developer is building a RAG-based Q&A bot with Amazon Bedrock Knowledge Bases. They need a managed vector store for document embeddings. Which service should they use?

Accepted Answer

Amazon OpenSearch Serverless. Amazon Bedrock Knowledge Bases requires a vector store to store and query document embeddings for Retrieval-Augmented Generation (RAG). Amazon OpenSearch Serverless provides a managed, scalable vector engine that supports k-NN (k-nearest neighbor) search, making it the correct choice for this use case. It integrates natively with Bedrock Knowledge Bases to handle embedding storage and similarity search without manual infrastructure management.

Answer

Amazon DynamoDB

Answer

Amazon RDS

Answer

Amazon S3

Question 12

Which pricing model does Amazon Bedrock use for foundation model inference?

Accepted Answer

Per-token. Amazon Bedrock charges for foundation model inference based on the number of tokens processed, which includes both input and output tokens. Each model has a specific per-token price, and the total cost is calculated by multiplying the token count by the model's rate. This is the standard pricing model for generative AI inference services like Bedrock.

Answer

Per-request

Answer

Per-hour instance

Answer

Per-GB storage

Question 13

A company wants to adapt a foundation model for a custom domain with very limited labeled data and minimal cost. Which approach is most suitable?

Accepted Answer

Prompt engineering with few-shot examples. Prompt engineering with few-shot examples is the most suitable approach because it allows the company to adapt a foundation model to a custom domain using very limited labeled data and minimal cost. By providing a few input-output examples directly in the prompt, the model can infer the desired task without any weight updates, making it efficient for low-resource scenarios.

Answer

Pre-training from scratch

Answer

Reinforcement learning from human feedback

Answer

Full fine-tuning

Question 14

A developer uses Amazon Bedrock to generate code. Some outputs contain syntax errors. What is the most likely cause?

Accepted Answer

The prompt lacks constraints or examples. Syntax errors in generated code typically arise when the prompt lacks sufficient constraints or examples to guide the model toward producing syntactically valid output. Amazon Bedrock's foundation models rely on clear instructions and few-shot examples to adhere to language syntax rules; without them, the model may generate plausible-looking but incorrect code.

Answer

The max_tokens is too low

Answer

The temperature is too high

Answer

The model lacks knowledge of the language

Question 15

Which AWS service provides a serverless API for accessing foundation models with per-token pricing?

Accepted Answer

Amazon Bedrock. Amazon Bedrock is a fully managed service that provides a serverless API for accessing foundation models (FMs) from providers like AI21 Labs, Anthropic, Cohere, Meta, and Stability AI. It offers per-token pricing, meaning you pay only for the number of tokens processed in both input and output, with no upfront commitments or infrastructure management required.

Answer

Amazon API Gateway

Answer

AWS Lambda

Answer

Amazon SageMaker