Amazon AWS Certified AI Practitioner AIF-C01 Practice Test 14
Question 130
An organization finds its LLM's response degraded when it stuffs thousands of irrelevant messages into the prompt. A colleague suggests: Keep adding irrelevant chat history because more text is always better. Which finding BEST explains why the suggestion should be rejected?
- Pretraining large general models requires substantial data and compute
- Irrelevant context increases cost and can distract or exceed token limits
- Higher temperature generally encourages variation, not stable repetitive outputs
- The shortcut is safe because its terminology is familiar to the team
Explanation
Correct Answer
B:Irrelevant context increases cost and can distract or exceed token limits. This is the direct defect in the proposed action: Keep adding irrelevant chat history because more text is always better.
Incorrect Answers
A:This concern belongs to another situation: A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot.
C:This finding is associated with a different problem: An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories.
D:Familiar wording is not proof that an action is appropriate. The actual defect is: Irrelevant context increases cost and can distract or exceed token limits.
Question 131
An organization finds its LLM's response degraded when it stuffs thousands of irrelevant messages into the prompt. After implementing Curate relevant context within the model's context limits, which observation BEST verifies the intended result?
- Shorter targeted context produces better grounded answers on a test set
- The team schedules a future test of whether shorter targeted context produces better grounded answers on a test set, but has no result
- The team approves Curate relevant context within the model's context limits but does not run an operational test
- Repeated tests show reduced wording variation while factual validation remains separate
Explanation
Correct Answer
A:The observed result, Shorter targeted context produces better grounded answers on a test set, provides direct evidence. Context engineering prioritizes useful instructions and evidence within a finite window.
Incorrect Answers
B:An unperformed future check is not evidence. Verify the actual outcome: Shorter targeted context produces better grounded answers on a test set.
C:The required observable condition is: Shorter targeted context produces better grounded answers on a test set.
D:This observation would be evidence for another case: An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories, not the current scenario.
Question 132
The initial scenario was: An organization finds its LLM's response degraded when it stuffs thousands of irrelevant messages into the prompt. The original control is already in place. Now users then require long policy documents without prompt overflow. Which ADDITIONAL action BEST addresses the new condition?
- Keep adding irrelevant chat history because more text is always better
- Curate relevant context within the model's context limits
- Retrieve only relevant policy chunks as needed
- Shorter targeted context produces better grounded answers on a test set
Explanation
Correct Answer
C:Users then require long policy documents without prompt overflow. The appropriate added action is: Retrieve only relevant policy chunks as needed.
Incorrect Answers
A:This original shortcut remains unsuitable: Irrelevant context increases cost and can distract or exceed token limits.
B:This was the original control and is already implemented. Context engineering prioritizes useful instructions and evidence within a finite window.
D:This describes evidence for the original approach, not a new action: Shorter targeted context produces better grounded answers on a test set.
Question 133
An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories. Which option BEST meets the stated AI/ML or AWS requirement?
- Start with a suitable pretrained FM to reduce data and training burden
- Increase temperature to maximize answer consistency
- Consider a higher temperature with quality review
- Use a lower temperature setting for more deterministic sampling tendencies
Explanation
Correct Answer
D:Lower sampling temperature generally reduces randomness in generated responses. In this situation, choose: Use a lower temperature setting for more deterministic sampling tendencies.
Incorrect Answers
A:This can be valid for a different task: A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot. The current request is different and needs the selected approach.
B:Higher temperature generally encourages variation, not stable repetitive outputs. The proposal, Increase temperature to maximize answer consistency, does not satisfy this need.
C:This is appropriate only after a distinct change: Marketing later requests diverse creative headline ideas. It is not the best initial response to an operator adjusts sampling so an ai assistant provides more repeatable policy answers rather than creative stories.
Question 134
An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories. A colleague suggests: Increase temperature to maximize answer consistency. Which finding BEST explains why the suggestion should be rejected?
- The shortcut is safe because its terminology is familiar to the team
- Higher temperature generally encourages variation, not stable repetitive outputs
- A textual claim is not evidence of an executed tool request
- Pretraining large general models requires substantial data and compute
Explanation
Correct Answer
B:Higher temperature generally encourages variation, not stable repetitive outputs. This is the direct defect in the proposed action: Increase temperature to maximize answer consistency.
Incorrect Answers
A:The actual defect is: Higher temperature generally encourages variation, not stable repetitive outputs.
C:This concern belongs to another situation: A chatbot periodically calls an approved order-status API after interpreting a customer's request.
D:This finding is associated with a different problem: A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot.
Question 135
An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories. After implementing Use a lower temperature setting for more deterministic sampling tendencies, which observation BEST verifies the intended result?
- A pilot reaches acceptable performance using an existing model without new pretraining
- Repeated tests show reduced wording variation while factual validation remains separate
- The team schedules a future test of whether repeated tests show reduced wording variation while factual validation remains separate, but has no result
- The team approves Use a lower temperature setting for more deterministic sampling tendencies but does not run an operational test
Explanation
Correct Answer
B:The observed result, Repeated tests show reduced wording variation while factual validation remains separate, provides direct evidence. Lower sampling temperature generally reduces randomness in generated responses.
Incorrect Answers
A:This observation would be evidence for another case: A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot, not the current scenario.
C:Verify the actual outcome: Repeated tests show reduced wording variation while factual validation remains separate.
D:The required observable condition is: Repeated tests show reduced wording variation while factual validation remains separate.
Question 136
The initial scenario was: An operator adjusts sampling so an AI assistant provides more repeatable policy answers rather than creative stories. The original control is already in place. Now marketing later requests diverse creative headline ideas. Which ADDITIONAL action BEST addresses the new condition?
- Use a lower temperature setting for more deterministic sampling tendencies
- Repeated tests show reduced wording variation while factual validation remains separate
- Consider a higher temperature with quality review
- Increase temperature to maximize answer consistency
Explanation
Correct Answer
C:Marketing later requests diverse creative headline ideas. The appropriate added action is: Consider a higher temperature with quality review.
Incorrect Answers
A:This was the original control and is already implemented. Lower sampling temperature generally reduces randomness in generated responses.
B:This describes evidence for the original approach, not a new action: Repeated tests show reduced wording variation while factual validation remains separate.
D:This original shortcut remains unsuitable: Higher temperature generally encourages variation, not stable repetitive outputs.
Question 137
A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot. Which option BEST meets the stated AI/ML or AWS requirement?
- Assume a custom foundation model can be pre-trained from one page of examples
- Describe agent tool use as controlled invocation of external capabilities
- Evaluate targeted customization or retrieval against domain examples
- Start with a suitable pretrained FM to reduce data and training burden
Explanation
Correct Answer
D:Pretraining a new foundation model is resource-intensive and usually unnecessary for a narrow pilot. In this situation, choose: Start with a suitable pretrained FM to reduce data and training burden.
Incorrect Answers
A:Pretraining large general models requires substantial data and compute. The proposal, Assume a custom foundation model can be pre-trained from one page of examples, does not satisfy this need.
B:This can be valid for a different task: A chatbot periodically calls an approved order-status API after interpreting a customer's request. The current request is different and needs the selected approach.
C:This is appropriate only after a distinct change: The model then needs better performance on specialized terminology. It is not the best initial response to a team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot.
Question 138
A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot. A colleague suggests: Assume a custom foundation model can be pre-trained from one page of examples. Which finding BEST explains why the suggestion should be rejected?
- A textual claim is not evidence of an executed tool request
- The shortcut is safe because its terminology is familiar to the team
- Pretraining large general models requires substantial data and compute
- A connection protocol does not remove independent tool and identity permissions
Explanation
Correct Answer
C:Pretraining large general models requires substantial data and compute. This is the direct defect in the proposed action: Assume a custom foundation model can be pre-trained from one page of examples.
Incorrect Answers
A:This finding is associated with a different problem: A chatbot periodically calls an approved order-status API after interpreting a customer's request.
B:Familiar wording is not proof that an action is appropriate. The actual defect is: Pretraining large general models requires substantial data and compute.
D:This concern belongs to another situation: A platform team wants a standard way for AI agents to discover tool functions exposed by multiple internal services.
Question 139
A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot. After implementing Start with a suitable pretrained FM to reduce data and training burden, which observation BEST verifies the intended result?
- A pilot reaches acceptable performance using an existing model without new pretraining
- The team schedules a future test of whether a pilot reaches acceptable performance using an existing model without new pretraining, but has no result
- The team approves Start with a suitable pretrained FM to reduce data and training burden but does not run an operational test
- An audit trail confirms actual authorized API calls and returned order status
Explanation
Correct Answer
A:The observed result, A pilot reaches acceptable performance using an existing model without new pretraining, provides direct evidence. Pretraining a new foundation model is resource-intensive and usually unnecessary for a narrow pilot.
Incorrect Answers
B:Verify the actual outcome: A pilot reaches acceptable performance using an existing model without new pretraining.
C:The required observable condition is: A pilot reaches acceptable performance using an existing model without new pretraining.
D:This observation would be evidence for another case: A chatbot periodically calls an approved order-status API after interpreting a customer's request, not the current scenario.
Question 140
The initial scenario was: A team compares training a foundation model from scratch with using an existing pre-trained model for a small internal pilot. The original control is already in place. Now the model then needs better performance on specialized terminology. Which ADDITIONAL action BEST addresses the new condition?
- Evaluate targeted customization or retrieval against domain examples
- Assume a custom foundation model can be pre-trained from one page of examples
- Start with a suitable pretrained FM to reduce data and training burden
- A pilot reaches acceptable performance using an existing model without new pretraining
Explanation
Correct Answer
A:The model then needs better performance on specialized terminology. The appropriate added action is: Evaluate targeted customization or retrieval against domain examples.
Incorrect Answers
B:This original shortcut remains unsuitable: Pretraining large general models requires substantial data and compute.
C:This was the original control and is already implemented. Pretraining a new foundation model is resource-intensive and usually unnecessary for a narrow pilot.
D:This describes evidence for the original approach, not a new action: A pilot reaches acceptable performance using an existing model without new pretraining.