Microsoft AI-102 (Designing and Implementing a Microsoft Azure AI Solution) Practice Test 23

Question 220

A vision model confidently reports a tiny serial number that cannot be read at the provided resolution. How should the result be handled?

  1. Validate tool arguments and user authorization before invoking the side-effecting operation
  2. Call an embedding model to produce numeric vectors for the index field
  3. Treat low-resolution visual transcription as uncertain and validate with OCR or higher-quality imagery
  4. Use a compatible model and its supported structured-output facility

Explanation

Correct Answer

C:

Model image understanding can miss small text; confident wording is not proof of exact character recognition. In the multimodal content-generation product, confirm the fix by retesting the affected behavior after treating low-resolution visual transcription as uncertain and validate with OCR or higher-quality imagery.

Incorrect Answers

A:

That change addresses another failure mode: a model's proposed tool arguments do not constitute proof that the user has permission to perform an action; it does not address this case, for which the relevant intervention is to treat low-resolution visual transcription as uncertain and validate with OCR or higher-quality imagery.

B:

Text responses are not automatically vector representations; vector indexes expect numeric arrays of configured dimensions in its own workflow, but that is a different correction from the present requirement to treat low-resolution visual transcription as uncertain and validate with OCR or higher-quality imagery. The reported issue would remain.

D:

Although useful elsewhere in the multimodal content-generation product, a general text model can generate prose, but reliable programmatic classification needs an enforceable response contract when available; the present case is different and calls for the team to treat low-resolution visual transcription as uncertain and validate with OCR or higher-quality imagery.

Question 221

A prompt revision improves ten cherry-picked samples but worsens common cases. What release gate should be used?

  1. Update and reindex the approved document corpus, including deletion or version invalidation
  2. Run a stable representative regression suite before promoting the prompt
  3. Instrument correlated retrieval, prompt construction, model call, and tool spans
  4. Apply a constrained review step that checks the draft against explicit output requirements

Explanation

Correct Answer

B:

A handful of successes cannot establish overall improvement or catch regressions across known scenarios. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after running a stable representative regression suite before promoting the prompt.

Incorrect Answers

A:

Grounding quality depends on current indexed evidence, not only model weights in its own workflow, but that is a different correction from the present requirement to run a stable representative regression suite before promoting the prompt. The reported issue would remain.

C:

Although useful elsewhere in the deployed enterprise chat and retrieval product, step-level traces attribute latency to actual pipeline stages and expose intermittent bottlenecks; the present case is different and calls for the team to run a stable representative regression suite before promoting the prompt.

D:

That change addresses another failure mode: a review pass can detect omissions, but must be evaluated rather than assumed to guarantee correctness; it does not address this case, for which the relevant intervention is to run a stable representative regression suite before promoting the prompt.

Question 222

Search retrieval improves after new documents arrive, but the assistant still cites old passages from a cache. What should be changed?

  1. Invalidate or version the retrieval cache when the indexed content changes
  2. Validate self-reviewed outputs against external ground truth and safety criteria
  3. Hold retrieval inputs and evaluation criteria consistent when isolating model differences
  4. Confirm production clients invoke the deployed fine-tuned model version

Explanation

Correct Answer

A:

A stale application cache can mask successful index updates and keep outdated evidence in prompts. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after invalidating or version the retrieval cache when the indexed content changes.

Incorrect Answers

B:

Although useful elsewhere in the deployed enterprise chat and retrieval product, a model can repeat or endorse its own mistakes; self-evaluation is not independent evidence; the present case is different and calls for the team to invalidate or version the retrieval cache when the indexed content changes.

C:

Comparable experiments change the intended variable while holding other significant conditions stable in its own workflow, but that is a different correction from the present requirement to invalidate or version the retrieval cache when the indexed content changes. The reported issue would remain.

D:

That change addresses another failure mode: training an adapted artifact does not alter a separate base-model deployment automatically; it does not address this case, for which the relevant intervention is to invalidate or version the retrieval cache when the indexed content changes.

Question 223

A RAG assistant answers from a policy rescinded last week. The model and prompt are unchanged. What should be refreshed?

  1. Run a stable representative regression suite before promoting the prompt
  2. Apply a constrained review step that checks the draft against explicit output requirements
  3. Update and reindex the approved document corpus, including deletion or version invalidation
  4. Instrument correlated retrieval, prompt construction, model call, and tool spans

Explanation

Correct Answer

C:

Grounding quality depends on current indexed evidence, not only model weights. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after updating and reindex the approved document corpus, including deletion or version invalidation.

Incorrect Answers

A:

Although useful elsewhere in the deployed enterprise chat and retrieval product, a handful of successes cannot establish overall improvement or catch regressions across known scenarios; the present case is different and calls for the team to update and reindex the approved document corpus, including deletion or version invalidation.

B:

A review pass can detect omissions, but must be evaluated rather than assumed to guarantee correctness in its own workflow, but that is a different correction from the present requirement to update and reindex the approved document corpus, including deletion or version invalidation. The reported issue would remain.

D:

That change addresses another failure mode: step-level traces attribute latency to actual pipeline stages and expose intermittent bottlenecks; it does not address this case, for which the relevant intervention is to update and reindex the approved document corpus, including deletion or version invalidation.

Question 224

A deployed enterprise chat and retrieval product must compare model or prompt changes against known expected behavior before releasing. Reviewers will accept the implementation when version comparisons expose quality regressions. Which approach best fulfills that requirement?

  1. Use a model reflection or self-review pass
  2. Run representative regression evaluations
  3. Fine-tune a supported language model
  4. Refresh retrieval grounding data

Explanation

Correct Answer

B:

Evaluations score responses against a sample dataset for qualities such as relevance, groundedness, or safety. For this deployed enterprise chat and retrieval product, the intended evidence is that version comparisons expose quality regressions. A test score does not enforce real-time content filtering on future requests.

Incorrect Answers

A:

A reflection step asks the model or a separate evaluator to critique and revise a generated draft according to stated checks; it may be used to assess a draft response against explicit criteria before presenting the final answer in a deployed enterprise chat and retrieval product, but the current requirement is instead to compare model or prompt changes against known expected behavior before releasing. That separate capability does not prove that version comparisons expose quality regressions; Self-critique may improve quality but does not guarantee truth or replace external evidence and evaluations, so selecting that alternative would not meet the stated acceptance check.

C:

Fine-tuning adjusts model behavior by training on examples rather than supplying the same context every turn; it may be used to adapt recurring domain-specific output style using approved example data and a supported tuning model in a deployed enterprise chat and retrieval product, but the current requirement is instead to compare model or prompt changes against known expected behavior before releasing. That separate capability does not prove that version comparisons expose quality regressions; It is not the best means of keeping rapidly changing company policies current, so selecting that alternative would not meet the stated acceptance check.

D:

Updating the indexed source corpus changes what the RAG retriever can supply on future queries; it may be used to make answers reflect updated internal policies without retraining the language model in a deployed enterprise chat and retrieval product, but the current requirement is instead to compare model or prompt changes against known expected behavior before releasing. That separate capability does not prove that version comparisons expose quality regressions; It does not change the underlying model weights or solve a malformed JSON schema by itself, so selecting that alternative would not meet the stated acceptance check.

Question 225

The solution brief for a deployed enterprise chat and retrieval product requires the system to assess a draft response against explicit criteria before presenting the final answer. A successful test confirms that a draft is reviewed against the acceptance rubric before release. What should the team use?

  1. Run representative regression evaluations
  2. Use a model reflection or self-review pass
  3. Fine-tune a supported language model
  4. Refresh retrieval grounding data

Explanation

Correct Answer

B:

A reflection step asks the model or a separate evaluator to critique and revise a generated draft according to stated checks. For this deployed enterprise chat and retrieval product, the intended evidence is that a draft is reviewed against the acceptance rubric before release. Self-critique may improve quality but does not guarantee truth or replace external evidence and evaluations.

Incorrect Answers

A:

Evaluations score responses against a sample dataset for qualities such as relevance, groundedness, or safety; it may be used to compare model or prompt changes against known expected behavior before releasing in a deployed enterprise chat and retrieval product, but the current requirement is instead to assess a draft response against explicit criteria before presenting the final answer. That separate capability does not prove that a draft is reviewed against the acceptance rubric before release; A test score does not enforce real-time content filtering on future requests, so selecting that alternative would not meet the stated acceptance check.

C:

Fine-tuning adjusts model behavior by training on examples rather than supplying the same context every turn; it may be used to adapt recurring domain-specific output style using approved example data and a supported tuning model in a deployed enterprise chat and retrieval product, but the current requirement is instead to assess a draft response against explicit criteria before presenting the final answer. That separate capability does not prove that a draft is reviewed against the acceptance rubric before release; It is not the best means of keeping rapidly changing company policies current, so selecting that alternative would not meet the stated acceptance check.

D:

Updating the indexed source corpus changes what the RAG retriever can supply on future queries; it may be used to make answers reflect updated internal policies without retraining the language model in a deployed enterprise chat and retrieval product, but the current requirement is instead to assess a draft response against explicit criteria before presenting the final answer. That separate capability does not prove that a draft is reviewed against the acceptance rubric before release; It does not change the underlying model weights or solve a malformed JSON schema by itself, so selecting that alternative would not meet the stated acceptance check.

Question 226

A firm wants its assistant to follow a consistent brand voice from hundreds of approved examples. Which decision precedes fine-tuning?

  1. Instrument correlated retrieval, prompt construction, model call, and tool spans
  2. Update and reindex the approved document corpus, including deletion or version invalidation
  3. Run a stable representative regression suite before promoting the prompt
  4. Verify that the chosen model supports tuning and prepare licensed high-quality example pairs

Explanation

Correct Answer

D:

Fine-tuning is not universally supported and requires curated examples appropriate to the target behavior. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after verifying that the chosen model supports tuning and prepare licensed high-quality example pairs.

Incorrect Answers

A:

That change addresses another failure mode: step-level traces attribute latency to actual pipeline stages and expose intermittent bottlenecks; it does not address this case, for which the relevant intervention is to verify that the chosen model supports tuning and prepare licensed high-quality example pairs.

B:

Grounding quality depends on current indexed evidence, not only model weights in its own workflow, but that is a different correction from the present requirement to verify that the chosen model supports tuning and prepare licensed high-quality example pairs. The reported issue would remain.

C:

Although useful elsewhere in the deployed enterprise chat and retrieval product, a handful of successes cannot establish overall improvement or catch regressions across known scenarios; the present case is different and calls for the team to verify that the chosen model supports tuning and prepare licensed high-quality example pairs.

Question 227

A release for a deployed enterprise chat and retrieval product is blocked until the application can make answers reflect updated internal policies without retraining the language model. The tester will verify that revised documents appear in subsequent grounded answers. What should be implemented?

  1. Run representative regression evaluations
  2. Refresh retrieval grounding data
  3. Use a model reflection or self-review pass
  4. Instrument prompt and tool traces

Explanation

Correct Answer

B:

Updating the indexed source corpus changes what the RAG retriever can supply on future queries. For this deployed enterprise chat and retrieval product, the intended evidence is that revised documents appear in subsequent grounded answers. It does not change the underlying model weights or solve a malformed JSON schema by itself.

Incorrect Answers

A:

Evaluations score responses against a sample dataset for qualities such as relevance, groundedness, or safety; it may be used to compare model or prompt changes against known expected behavior before releasing in a deployed enterprise chat and retrieval product, but the current requirement is instead to make answers reflect updated internal policies without retraining the language model. That separate capability does not prove that revised documents appear in subsequent grounded answers; A test score does not enforce real-time content filtering on future requests, so selecting that alternative would not meet the stated acceptance check.

C:

A reflection step asks the model or a separate evaluator to critique and revise a generated draft according to stated checks; it may be used to assess a draft response against explicit criteria before presenting the final answer in a deployed enterprise chat and retrieval product, but the current requirement is instead to make answers reflect updated internal policies without retraining the language model. That separate capability does not prove that revised documents appear in subsequent grounded answers; Self-critique may improve quality but does not guarantee truth or replace external evidence and evaluations, so selecting that alternative would not meet the stated acceptance check.

D:

Tracing records model calls, tool operations, and relevant workflow spans when configured; it may be used to diagnose latency and incorrect tool sequences during production conversations in a deployed enterprise chat and retrieval product, but the current requirement is instead to make answers reflect updated internal policies without retraining the language model. That separate capability does not prove that revised documents appear in subsequent grounded answers; It does not itself decide whether the model output is safe or accurate, so selecting that alternative would not meet the stated acceptance check.

Question 228

A model's initial answer may omit an important safety disclaimer. A second independent pass is planned. How should it be used?

  1. Instrument correlated retrieval, prompt construction, model call, and tool spans
  2. Run a stable representative regression suite before promoting the prompt
  3. Verify that the chosen model supports tuning and prepare licensed high-quality example pairs
  4. Apply a constrained review step that checks the draft against explicit output requirements

Explanation

Correct Answer

D:

A review pass can detect omissions, but must be evaluated rather than assumed to guarantee correctness. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after applying a constrained review step that checks the draft against explicit output requirements.

Incorrect Answers

A:

Although useful elsewhere in the deployed enterprise chat and retrieval product, step-level traces attribute latency to actual pipeline stages and expose intermittent bottlenecks; the present case is different and calls for the team to apply a constrained review step that checks the draft against explicit output requirements.

B:

That change addresses another failure mode: a handful of successes cannot establish overall improvement or catch regressions across known scenarios; it does not address this case, for which the relevant intervention is to apply a constrained review step that checks the draft against explicit output requirements.

C:

Fine-tuning is not universally supported and requires curated examples appropriate to the target behavior in its own workflow, but that is a different correction from the present requirement to apply a constrained review step that checks the draft against explicit output requirements. The reported issue would remain.

Question 229

A retrieval system shows excellent matching on a test set containing only training passages. What should be tested instead?

  1. Evaluate with held-out queries and documents that represent production retrieval failures
  2. Trigger review selectively for outputs whose risk or complexity warrants its cost
  3. Retrieve the current authoritative price catalog at inference time
  4. Assign reliable correlation identifiers at request and conversation boundaries

Explanation

Correct Answer

A:

Realistic independent examples better measure generalization and avoid inflated retrieval metrics. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after evaluating with held-out queries and documents that represent production retrieval failures.

Incorrect Answers

B:

Conditional verification can control latency and spend while retaining checks on high-impact content in its own workflow, but that is a different correction from the present requirement to evaluate with held-out queries and documents that represent production retrieval failures. The reported issue would remain.

C:

Although useful elsewhere in the deployed enterprise chat and retrieval product, parameter updates are inefficient and unreliable for frequently changing source facts compared with live grounding; the present case is different and calls for the team to evaluate with held-out queries and documents that represent production retrieval failures.

D:

That change addresses another failure mode: correct trace correlation is required to reconstruct individual agent executions and avoid mixing users' actions; it does not address this case, for which the relevant intervention is to evaluate with held-out queries and documents that represent production retrieval failures.

Question 230

An engineering team assumes model self-review eliminates the need for test cases. What is missing?

  1. Validate self-reviewed outputs against external ground truth and safety criteria
  2. Confirm production clients invoke the deployed fine-tuned model version
  3. Invalidate or version the retrieval cache when the indexed content changes
  4. Hold retrieval inputs and evaluation criteria consistent when isolating model differences

Explanation

Correct Answer

A:

A model can repeat or endorse its own mistakes; self-evaluation is not independent evidence. In the deployed enterprise chat and retrieval product, confirm the fix by retesting the affected behavior after validating self-reviewed outputs against external ground truth and safety criteria.

Incorrect Answers

B:

Training an adapted artifact does not alter a separate base-model deployment automatically in its own workflow, but that is a different correction from the present requirement to validate self-reviewed outputs against external ground truth and safety criteria. The reported issue would remain.

C:

Although useful elsewhere in the deployed enterprise chat and retrieval product, a stale application cache can mask successful index updates and keep outdated evidence in prompts; the present case is different and calls for the team to validate self-reviewed outputs against external ground truth and safety criteria.

D:

That change addresses another failure mode: comparable experiments change the intended variable while holding other significant conditions stable; it does not address this case, for which the relevant intervention is to validate self-reviewed outputs against external ground truth and safety criteria.

Designing and Implementing a Microsoft Azure AI Solution (AI-102) Practice Tests