Amazon AWS AI Practitioner AIF-C01 Foundation Model Training Fine Tuning And Data Preparation Practice Test

 

AIF-C01 skills 3.3 | 26 original questions

This AWS Certified AI Practitioner AIF-C01 practice test focuses on foundation model training fine tuning and data preparation through original scenario-based questions aligned to AWS Exam Guide version 1.1 published April 30, 2026. Use the full ExamSnap AIF-C01 collection for broader practice across all five current exam domains. For broader exam preparation, review the Amazon AWS Certified AI Practitioner AIF-C01 Exam Dumps page.

Instructions: Select the best answer for each question. Review the rationale after answering. Each distractor includes a brief explanation of why it is not the strongest fit for the stated scenario.

Question 1

A workshop at Proseware Services focuses on a single decision: how to include the users, domains, languages, and edge cases the tuned model is expected to serve. Which option should the business analyst recommend? Stakeholders have ruled out a broad redesign and want the choice that most precisely addresses the stated need. The team is comparing 3 candidate designs after a 697-day proof of concept.

1. Representative dataset
2. Sufficient data volume
3. Reinforcement learning from human feedback (RLHF)
4. Inclusive dataset
5. Balanced dataset

Correct answer: A

Why: Representativeness reduces blind spots and improves generalization to the target population. It directly addresses the requirement in this scenario.

Option review:

A: Representativeness reduces blind spots and improves generalization to the target population. It directly addresses the requirement in this scenario.

B: Training-set size should be adequate for the adaptation method and task complexity. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: RLHF uses human judgments to provide reward or preference information during model alignment. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Balance can improve fairness and make performance metrics more representative. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Representative dataset – Representativeness reduces blind spots and improves generalization to the target population.

Question 2

For the fraud-review pilot at Lucerne Publishing, stakeholders need to continue broad language/model training on additional domain data before downstream fine-tuning or deployment. Which concept, service, or technique most directly addresses this goal? Operational ownership is already assigned, so the team is comparing technical fit rather than staffing models. The control owner requires evidence from 8 test groups before the 734-day release review.

1. Fine-tuning
2. Continuous pre-training
3. Pre-training
4. Distillation
5. Transfer learning

Correct answer: B

Why: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

Option review:

A: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

C: Pre-training establishes the base capabilities of a foundation model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Continuous pre-training – Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point.

Question 3

Lamna Healthcare is comparing alternatives for its analytics modernization. The business analyst needs to reuse knowledge learned in a source task or model and adapt it to a related target task. Which option is most appropriate while trying to control recurring cost? Existing application interfaces can accommodate any of the listed choices, so functional fit is the deciding factor. The project has 5 downstream consumers and a monthly review of approximately 771 sampled interactions.

1. Distillation
2. Instruction tuning
3. Transfer learning
4. Pre-training
5. Continuous pre-training

Correct answer: C

Why: Transfer learning reduces the need to learn everything from scratch. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Transfer learning reduces the need to learn everything from scratch. It directly addresses the requirement in this scenario.

D: Pre-training establishes the base capabilities of a foundation model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Transfer learning – Transfer learning reduces the need to learn everything from scratch.

Question 4

An architecture review at Contoso Retail has narrowed a compliance-assistant prototype decision to one requirement: remove low-quality, irrelevant, duplicated, or unsafe examples before fine-tuning. What should the data scientist select? Assume the required AWS capabilities are available in the selected Region and normal governance controls are in place. The rollout spans 2 application teams, each using the same approved requirement set for the next 808 days.

1. Diverse dataset
2. Sufficient data volume
3. Inclusive dataset
4. Data curation
5. Reinforcement learning from human feedback (RLHF)

Correct answer: D

Why: Curated data improves signal quality and reduces the chance of teaching undesirable behavior. It directly addresses the requirement in this scenario.

Option review:

A: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Training-set size should be adequate for the adaptation method and task complexity. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Curated data improves signal quality and reduces the chance of teaching undesirable behavior. It directly addresses the requirement in this scenario.

E: RLHF uses human judgments to provide reward or preference information during model alignment. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data curation – Curated data improves signal quality and reduces the chance of teaching undesirable behavior.

Question 5

The business analyst at Fourth Coffee is preparing a recommendation for a forecasting initiative. The recommendation must learn broad representations from large datasets before task-specific adaptation. Which choice is the best match? The review committee wants a direct mapping from the requirement to the chosen capability. The evaluation set contains examples from 7 business workflows and 845 recent production cases.

1. Fine-tuning
2. Instruction tuning
3. Distillation
4. Retrieval Augmented Generation (RAG)
5. Pre-training

Correct answer: E

Why: Pre-training establishes the base capabilities of a foundation model. It directly addresses the requirement in this scenario.

Option review:

A: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Pre-training establishes the base capabilities of a foundation model. It directly addresses the requirement in this scenario.

Learning point: Pre-training – Pre-training establishes the base capabilities of a foundation model.

Question 6

Margie Travel has completed discovery for a customer-support modernization. Before implementation, the data scientist must decide how to fine-tune on instruction-response examples to improve how the model follows task directions. Which choice best satisfies that requirement? The solution will serve multiple internal teams, so the recommendation should be reusable without changing the core requirement. The initial rollout covers 882 internal users across 4 business units.

1. Instruction tuning
2. Distillation
3. Continuous pre-training
4. Domain adaptation
5. Fine-tuning

Correct answer: A

Why: Instruction tuning teaches a model to respond appropriately to explicit instructions. It directly addresses the requirement in this scenario.

Option review:

A: Instruction tuning teaches a model to respond appropriately to explicit instructions. It directly addresses the requirement in this scenario.

B: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Instruction tuning – Instruction tuning teaches a model to respond appropriately to explicit instructions.

Question 7

While planning a agentic workflow trial, School of Fine Art identifies this requirement: verify permissions, lineage, retention, privacy, and approved use of fine-tuning data. Which option should the business analyst prioritize if the goal is to use current managed AWS capabilities? The decision must follow the workload characteristics rather than a preference for the largest model or newest service. The workload processes about 919 requests during its busiest hour and has a documented fallback path.

1. Balanced dataset
2. Data governance
3. Correct labeling
4. Curated data source
5. Data curation

Correct answer: B

Why: Governance requirements apply to training data as well as production inputs. It directly addresses the requirement in this scenario.

Option review:

A: Balance can improve fairness and make performance metrics more representative. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Governance requirements apply to training data as well as production inputs. It directly addresses the requirement in this scenario.

C: Label quality directly affects what the model learns. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Curation improves quality and supports governance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Curated data improves signal quality and reduces the chance of teaching undesirable behavior. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data governance – Governance requirements apply to training data as well as production inputs.

Question 8

A proof of concept at Northwind Analytics exposed a design decision for the data scientist: the solution must continue training a pretrained model on curated examples for a target behavior. Which option most directly solves that problem? The security baseline is already defined; the decision here concerns the specific capability described in the requirement. The pilot uses 956 representative records from 6 approved data sources.

1. Retrieval Augmented Generation (RAG)
2. Instruction tuning
3. Fine-tuning
4. Domain adaptation
5. Model distillation

Correct answer: C

Why: Fine-tuning modifies model weights to specialize performance. It directly addresses the requirement in this scenario.

Option review:

A: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Fine-tuning modifies model weights to specialize performance. It directly addresses the requirement in this scenario.

D: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Distillation can retain useful behavior in a smaller model while lowering serving cost. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Fine-tuning – Fine-tuning modifies model weights to specialize performance.

Question 9

Litware Financial is documenting the target state for a operations automation program. The business analyst needs a solution that can fine-tune or continue training with representative domain data to improve specialized terminology and behavior. Which option is the strongest fit? The recommendation must solve the stated requirement without introducing unrelated platform complexity. The first release supports 3 departments and is reviewed every 993 days.

1. Distillation
2. Fine-tuning
3. Continuous pre-training
4. Domain adaptation
5. Pre-training

Correct answer: D

Why: Domain adaptation specializes a general model for a target field. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Domain adaptation specializes a general model for a target field. It directly addresses the requirement in this scenario.

E: Pre-training establishes the base capabilities of a foundation model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Domain adaptation – Domain adaptation specializes a general model for a target field.

Question 10

1. Datum Research is reviewing a sales-assistant rollout. The data scientist has one primary requirement: use human preference signals to shape model behavior toward preferred responses. Which choice best fits the requirement? The design must remain supportable after launch, but no additional feature is required beyond the stated need. The service has a 70-millisecond internal response target for the affected workflow.
2. Balanced dataset
3. Data governance
4. Representative dataset
5. Inclusive dataset
6. Reinforcement learning from human feedback (RLHF)

Correct answer: E

Why: RLHF uses human judgments to provide reward or preference information during model alignment. It directly addresses the requirement in this scenario.

Option review:

A: Balance can improve fairness and make performance metrics more representative. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Governance requirements apply to training data as well as production inputs. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Representativeness reduces blind spots and improves generalization to the target population. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: RLHF uses human judgments to provide reward or preference information during model alignment. It directly addresses the requirement in this scenario.

Learning point: Reinforcement learning from human feedback (RLHF) – RLHF uses human judgments to provide reward or preference information during model alignment.

Question 11

During a design review for Coho Winery, the business analyst must train a smaller student model to reproduce useful behavior from a larger teacher model. The team also wants to keep the design easy to explain. What should the team choose? A short pilot window means the team prefers an approach that can be evaluated with clear success criteria. The team is comparing 5 candidate designs after a 107-day proof of concept.

1. Distillation
2. Instruction tuning
3. Continuous pre-training
4. Domain adaptation
5. Fine-tuning

Correct answer: A

Why: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. It directly addresses the requirement in this scenario.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Distillation – Distillation aims to reduce serving cost and latency while retaining much of the teacher capability.

Question 12

Lucerne Retail is moving a document-intelligence project from pilot to production. The key decision is how to extend model pre-training on additional unlabeled domain data before task-specific tuning. Which option is the strongest fit if the team wants to reduce manual handling? The architecture board will reject a choice that addresses a different problem from the one described. The control owner requires evidence from 2 test groups before the 144-day release review.

1. Distillation
2. Continuous pre-training
3. Transfer learning
4. Domain adaptation
5. Instruction tuning

Correct answer: B

Why: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. It directly addresses the requirement in this scenario.

C: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Continuous pre-training – Continuous pre-training is useful when substantial domain language or knowledge must be incorporated.

Question 13

A workshop at Tailspin Toys focuses on a single decision: how to ensure instruction-response pairs or target labels are accurate and consistent. Which option should the business analyst recommend? Budget has been approved for the project, but the team still wants to avoid unnecessary recurring consumption. The project has 7 downstream consumers and a monthly review of approximately 181 sampled interactions.

1. Reinforcement learning from human feedback (RLHF)
2. Diverse dataset
3. Correct labeling
4. Balanced dataset
5. Inclusive dataset

Correct answer: C

Why: Label quality directly affects what the model learns. It directly addresses the requirement in this scenario.

Option review:

A: RLHF uses human judgments to provide reward or preference information during model alignment. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Label quality directly affects what the model learns. It directly addresses the requirement in this scenario.

D: Balance can improve fairness and make performance metrics more representative. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Correct labeling – Label quality directly affects what the model learns.

Question 14

For the claims-processing redesign at City Power and Light, stakeholders need to continue broad language/model training on additional domain data before downstream fine-tuning or deployment. Which concept, service, or technique most directly addresses this goal? The team will validate the result with representative production examples before rollout. The rollout spans 4 application teams, each using the same approved requirement set for the next 218 days.

1. Distillation
2. Instruction tuning
3. Pre-training from scratch
4. Continuous pre-training
5. Retrieval Augmented Generation (RAG)

Correct answer: D

Why: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Pre-training is the most resource-intensive customization path and is rarely justified for ordinary application adaptation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

E: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Continuous pre-training – Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point.

Question 15

Consolidated Messenger is comparing alternatives for its personalization program. The business analyst needs to reuse knowledge learned in a source task or model and adapt it to a related target task. Which option is most appropriate while trying to control recurring cost? The pilot has representative data, and the team will measure the selected approach against an agreed acceptance threshold. The evaluation set contains examples from 9 business workflows and 255 recent production cases.

1. Continuous pre-training
2. Pre-training
3. Fine-tuning
4. Instruction tuning
5. Transfer learning

Correct answer: E

Why: Transfer learning reduces the need to learn everything from scratch. It directly addresses the requirement in this scenario.

Option review:

A: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Pre-training establishes the base capabilities of a foundation model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Transfer learning reduces the need to learn everything from scratch. It directly addresses the requirement in this scenario.

Learning point: Transfer learning – Transfer learning reduces the need to learn everything from scratch.

Question 16

An architecture review at Nod Publishers has narrowed a developer-productivity pilot decision to one requirement: provide enough diverse examples to learn the target behavior without relying on a handful of anecdotes. What should the data scientist select? The team will document the rationale for auditors and wants the recommendation to be defensible from the scenario facts. The initial rollout covers 292 internal users across 6 business units.

1. Sufficient data volume
2. Balanced dataset
3. Data governance
4. Diverse dataset
5. Inclusive dataset

Correct answer: A

Why: Training-set size should be adequate for the adaptation method and task complexity. It directly addresses the requirement in this scenario.

Option review:

A: Training-set size should be adequate for the adaptation method and task complexity. It directly addresses the requirement in this scenario.

B: Balance can improve fairness and make performance metrics more representative. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Governance requirements apply to training data as well as production inputs. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Sufficient data volume – Training-set size should be adequate for the adaptation method and task complexity.

Question 17

The business analyst at Fabrikam Health is preparing a recommendation for a fraud-review pilot. The recommendation must continue training a pretrained model on curated examples for a target behavior. Which choice is the best match? The team wants the least complex technically correct choice that satisfies the requirement. The workload processes about 329 requests during its busiest hour and has a documented fallback path.

1. Transfer learning
2. Fine-tuning
3. Pre-training from scratch
4. Retrieval Augmented Generation (RAG)
5. In-context learning

Correct answer: B

Why: Fine-tuning modifies model weights to specialize performance. It directly addresses the requirement in this scenario.

Option review:

A: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Fine-tuning modifies model weights to specialize performance. It directly addresses the requirement in this scenario.

C: Pre-training is the most resource-intensive customization path and is rarely justified for ordinary application adaptation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: In-context learning is fast to iterate but consumes context tokens on each request. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Fine-tuning – Fine-tuning modifies model weights to specialize performance.

Question 18

Wingtip Logistics has completed discovery for a analytics modernization. Before implementation, the data scientist must decide how to fine-tune or continue training with representative domain data to improve specialized terminology and behavior. Which choice best satisfies that requirement? The workload has passed basic feasibility checks, so the remaining question is which approach best matches the requirement. The pilot uses 366 representative records from 8 approved data sources.

1. Transfer learning
2. Fine-tuning
3. Domain adaptation
4. Instruction tuning
5. Continuous pre-training

Correct answer: C

Why: Domain adaptation specializes a general model for a target field. It directly addresses the requirement in this scenario.

Option review:

A: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Domain adaptation specializes a general model for a target field. It directly addresses the requirement in this scenario.

D: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Domain adaptation – Domain adaptation specializes a general model for a target field.

Question 19

While planning a compliance-assistant prototype, Trey Research identifies this requirement: include the users, domains, languages, and edge cases the tuned model is expected to serve. Which option should the business analyst prioritize if the goal is to use current managed AWS capabilities? Stakeholders have ruled out a broad redesign and want the choice that most precisely addresses the stated need. The first release supports 5 departments and is reviewed every 403 days.

1. Diverse dataset
2. Data governance
3. Curated data source
4. Representative dataset
5. Reinforcement learning from human feedback (RLHF)

Correct answer: D

Why: Representativeness reduces blind spots and improves generalization to the target population. It directly addresses the requirement in this scenario.

Option review:

A: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Governance requirements apply to training data as well as production inputs. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Curation improves quality and supports governance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Representativeness reduces blind spots and improves generalization to the target population. It directly addresses the requirement in this scenario.

E: RLHF uses human judgments to provide reward or preference information during model alignment. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Representative dataset – Representativeness reduces blind spots and improves generalization to the target population.

Question 20

A proof of concept at Bellows College exposed a design decision for the data scientist: the solution must learn broad representations from large datasets before task-specific adaptation. Which option most directly solves that problem? Operational ownership is already assigned, so the team is comparing technical fit rather than staffing models. The service has a 440-millisecond internal response target for the affected workflow.

1. Domain adaptation
2. Instruction tuning
3. Fine-tuning
4. In-context learning
5. Pre-training

Correct answer: E

Why: Pre-training establishes the base capabilities of a foundation model. It directly addresses the requirement in this scenario.

Option review:

A: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: In-context learning is fast to iterate but consumes context tokens on each request. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Pre-training establishes the base capabilities of a foundation model. It directly addresses the requirement in this scenario.

Learning point: Pre-training – Pre-training establishes the base capabilities of a foundation model.

Question 21

Blue Yonder Airlines is documenting the target state for a customer-support modernization. The business analyst needs a solution that can fine-tune on instruction-response examples to improve how the model follows task directions. Which option is the strongest fit? Existing application interfaces can accommodate any of the listed choices, so functional fit is the deciding factor. The team is comparing 7 candidate designs after a 477-day proof of concept.

1. Instruction tuning
2. Distillation
3. Fine-tuning
4. Transfer learning
5. Pre-training

Correct answer: A

Why: Instruction tuning teaches a model to respond appropriately to explicit instructions. It directly addresses the requirement in this scenario.

Option review:

A: Instruction tuning teaches a model to respond appropriately to explicit instructions. It directly addresses the requirement in this scenario.

B: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Pre-training establishes the base capabilities of a foundation model. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Instruction tuning – Instruction tuning teaches a model to respond appropriately to explicit instructions.

Question 22

Woodgrove Bank is reviewing a agentic workflow trial. The data scientist has one primary requirement: verify permissions, lineage, retention, privacy, and approved use of fine-tuning data. Which choice best fits the requirement? Assume the required AWS capabilities are available in the selected Region and normal governance controls are in place. The control owner requires evidence from 4 test groups before the 514-day release review.

1. Sufficient data volume
2. Data governance
3. Data curation
4. Correct labeling
5. Diverse dataset

Correct answer: B

Why: Governance requirements apply to training data as well as production inputs. It directly addresses the requirement in this scenario.

Option review:

A: Training-set size should be adequate for the adaptation method and task complexity. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Governance requirements apply to training data as well as production inputs. It directly addresses the requirement in this scenario.

C: Curated data improves signal quality and reduces the chance of teaching undesirable behavior. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Label quality directly affects what the model learns. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Data governance – Governance requirements apply to training data as well as production inputs.

Question 23

During a design review for Wide World Importers, the business analyst must train a smaller student model to reproduce useful behavior from a larger teacher model. The team also wants to keep the design easy to explain. What should the team choose? The review committee wants a direct mapping from the requirement to the chosen capability. The project has 9 downstream consumers and a monthly review of approximately 551 sampled interactions.

1. Continuous pre-training
2. Model distillation
3. Distillation
4. Retrieval Augmented Generation (RAG)
5. Instruction tuning

Correct answer: C

Why: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. It directly addresses the requirement in this scenario.

Option review:

A: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Distillation can retain useful behavior in a smaller model while lowering serving cost. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. It directly addresses the requirement in this scenario.

D: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Distillation – Distillation aims to reduce serving cost and latency while retaining much of the teacher capability.

Question 24

VanArsdel Media is moving a operations automation program from pilot to production. The key decision is how to extend model pre-training on additional unlabeled domain data before task-specific tuning. Which option is the strongest fit if the team wants to reduce manual handling? The solution will serve multiple internal teams, so the recommendation should be reusable without changing the core requirement. The rollout spans 6 application teams, each using the same approved requirement set for the next 588 days.

1. Distillation
2. Instruction tuning
3. Transfer learning
4. Continuous pre-training
5. Domain adaptation

Correct answer: D

Why: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. It directly addresses the requirement in this scenario.

Option review:

A: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Transfer learning reduces the need to learn everything from scratch. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Continuous pre-training is useful when substantial domain language or knowledge must be incorporated. It directly addresses the requirement in this scenario.

E: Domain adaptation specializes a general model for a target field. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Continuous pre-training – Continuous pre-training is useful when substantial domain language or knowledge must be incorporated.

Question 25

A workshop at Datum Dynamics focuses on a single decision: how to provide enough diverse examples to learn the target behavior without relying on a handful of anecdotes. Which option should the business analyst recommend? The decision must follow the workload characteristics rather than a preference for the largest model or newest service. The evaluation set contains examples from 3 business workflows and 625 recent production cases.

1. Diverse dataset
2. Inclusive dataset
3. Representative dataset
4. Reinforcement learning from human feedback (RLHF)
5. Sufficient data volume

Correct answer: E

Why: Training-set size should be adequate for the adaptation method and task complexity. It directly addresses the requirement in this scenario.

Option review:

A: Diversity improves coverage of real-world variation. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

B: Inclusive data helps reduce systematic blind spots. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Representativeness reduces blind spots and improves generalization to the target population. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: RLHF uses human judgments to provide reward or preference information during model alignment. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: Training-set size should be adequate for the adaptation method and task complexity. It directly addresses the requirement in this scenario.

Learning point: Sufficient data volume – Training-set size should be adequate for the adaptation method and task complexity.

Question 26

For the internal search upgrade at Alpine Ski House, stakeholders need to continue broad language/model training on additional domain data before downstream fine-tuning or deployment. Which concept, service, or technique most directly addresses this goal? The security baseline is already defined; the decision here concerns the specific capability described in the requirement. The initial rollout covers 662 internal users across 8 business units.

1. Continuous pre-training
2. Distillation
3. Instruction tuning
4. Fine-tuning
5. Retrieval Augmented Generation (RAG)

Correct answer: A

Why: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

Option review:

A: Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point. It directly addresses the requirement in this scenario.

B: Distillation aims to reduce serving cost and latency while retaining much of the teacher capability. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

C: Instruction tuning teaches a model to respond appropriately to explicit instructions. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

D: Fine-tuning modifies model weights to specialize performance. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

E: RAG is attractive for changing factual knowledge and source attribution but adds retrieval infrastructure and latency. This can be appropriate in another scenario, but it does not most directly satisfy the requirement described here.

Learning point: Continuous pre-training – Continuous pre-training can deepen domain knowledge while retaining a general pretrained starting point.