Microsoft AI-103 (Developing AI Apps and Agents on Azure) Practice Test 18

Question 170

A cloud engineering team must answer questions about changing internal procedures with source-backed responses. What is the most appropriate Azure AI implementation?

  1. Use a multimodal request for document-plus-image reasoning
  2. Use grounded retrieval-augmented generation
  3. Route conditional branches using rules
  4. Use constrained structured JSON output

Explanation

Correct Answer

B:

For the stated requirement to answer questions about changing internal procedures with source-backed responses, RAG retrieves relevant evidence at runtime and gives it to a generative model; model pretraining alone cannot guarantee current policy text; this is the mechanism that the system must implement.

Incorrect Answers

A:

For a system whose objective is to reason about a photo and related user instructions together in one supported inference flow, a multimodal model can interpret visual input alongside textual context; relying on OCR alone loses non-textual relationships; this does not directly meet the required outcome to answer questions about changing internal procedures with source-backed responses.

C:

For a system whose objective is to choose among distinct business flows based on a validated account status, rules engines offer predictable decisions for simple structured conditions, while language models handle unstructured interpretation; this does not directly meet the required outcome to answer questions about changing internal procedures with source-backed responses.

D:

For a system whose objective is to send model-produced classifications into a strict downstream API contract, schema-constrained generation helps enforce machine-readable fields and types; application-level validation is still needed for business rules; this does not directly meet the required outcome to answer questions about changing internal procedures with source-backed responses.

Question 171

The implementation does not meet its business requirement: A response parser fails because the model emits a prose preface and changes JSON key names unpredictably. Which corrective action is appropriate?

  1. Reject invalid parameters with application-side validation before the business API call
  2. Provide image content to a vision-capable model with a bounded question and verification rubric
  3. Request structured outputs conforming to the expected schema and validate before ingestion
  4. Retrieve current authorized passages and cite them when forming the response

Explanation

Correct Answer

C:

The fault signal is that a response parser fails because the model emits a prose preface and changes JSON key names unpredictably; the targeted fix works because schema-constrained generation helps enforce machine-readable fields and types; application-level validation is still needed for business rules, restoring the ability to send model-produced classifications into a strict downstream API contract.

Incorrect Answers

A:

If the failing function were to prevent malformed model-generated parameters from reaching a write-capable API, this change could be relevant because tool schemas guide argument shape, but the application must check types, constraints, identity and authorization before execution; the actual symptom is that a response parser fails because the model emits a prose preface and changes JSON key names unpredictably, which calls for a different repair.

B:

If the failing function were to reason about a photo and related user instructions together in one supported inference flow, this change could be relevant because a multimodal model can interpret visual input alongside textual context; relying on OCR alone loses non-textual relationships; the actual symptom is that a response parser fails because the model emits a prose preface and changes JSON key names unpredictably, which calls for a different repair.

D:

If the failing function were to answer questions about changing internal procedures with source-backed responses, this change could be relevant because RAG retrieves relevant evidence at runtime and gives it to a generative model; model pretraining alone cannot guarantee current policy text; the actual symptom is that a response parser fails because the model emits a prose preface and changes JSON key names unpredictably, which calls for a different repair.

Question 172

During a regression test, the following unexpected result occurs: A question outside the indexed manuals receives an invented policy number instead of a request for clarification. What should the engineer do?

  1. Filter authorized indexed chunks using tenant, department, and effective-date fields
  2. Detect empty or low-confidence retrieval and return a supported no-answer or escalation response
  3. Attach retrieved chunk references to the claims they substantiate and validate that mapping
  4. Retrieve current authorized passages and cite them when forming the response

Explanation

Correct Answer

B:

The fault signal is that a question outside the indexed manuals receives an invented policy number instead of a request for clarification; the targeted fix works because an evidence threshold and abstention path distinguish unknown facts from model completion skill; confident text is not proof of source support, restoring the ability to avoid fabricating an answer to a question not covered by the organization's knowledge collection.

Incorrect Answers

A:

If the failing function were to ensure answers include only documents for a user's department and effective policy year, this change could be relevant because retrieval filters narrow the evidence set by structured attributes before the model sees it; prompt instructions are not security filters; the actual symptom is that a question outside the indexed manuals receives an invented policy number instead of a request for clarification, which calls for a different repair.

C:

If the failing function were to let a compliance reviewer locate the specific evidence supporting each generated assertion, this change could be relevant because citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; the actual symptom is that a question outside the indexed manuals receives an invented policy number instead of a request for clarification, which calls for a different repair.

D:

If the failing function were to answer questions about changing internal procedures with source-backed responses, this change could be relevant because RAG retrieves relevant evidence at runtime and gives it to a generative model; model pretraining alone cannot guarantee current policy text; the actual symptom is that a question outside the indexed manuals receives an invented policy number instead of a request for clarification, which calls for a different repair.

Question 173

During a regression test, the following unexpected result occurs: The user cancels an expensive request, but the server keeps receiving and billing for output. What should the engineer do?

  1. Propagate cancellation to the model request and close the stream on disconnect
  2. Reject invalid parameters with application-side validation before the business API call
  3. Attach retrieved chunk references to the claims they substantiate and validate that mapping
  4. Configure a credential chain using the workload identity and least-privilege role grants

Explanation

Correct Answer

A:

The fault signal is that the user cancels an expensive request, but the server keeps receiving and billing for output; the targeted fix works because streaming delivers incremental output; cancellation and backpressure prevent orphaned inference work and stale interface content, restoring the ability to show partial assistant output to a user while stopping work after the request is canceled.

Incorrect Answers

B:

If the failing function were to prevent malformed model-generated parameters from reaching a write-capable API, this change could be relevant because tool schemas guide argument shape, but the application must check types, constraints, identity and authorization before execution; the actual symptom is that the user cancels an expensive request, but the server keeps receiving and billing for output, which calls for a different repair.

C:

If the failing function were to let a compliance reviewer locate the specific evidence supporting each generated assertion, this change could be relevant because citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; the actual symptom is that the user cancels an expensive request, but the server keeps receiving and billing for output, which calls for a different repair.

D:

If the failing function were to authenticate a server-side generative app through Entra where supported, this change could be relevant because a workload identity and appropriate RBAC reduce static credential exposure while preserving auditable authorization decisions; the actual symptom is that the user cancels an expensive request, but the server keeps receiving and billing for output, which calls for a different repair.

Question 174

A release reviewer is assessing a feature intended to resolve deployment and endpoint from runtime configuration. Which test is most meaningful?

  1. Inspect request-target metadata and run an environment-isolation smoke test
  2. Test boundary values and confirm no rejected argument causes a side effect
  3. Measure cancellation propagation and discarded tokens after client disconnects
  4. Count handled failures and confirm consistent error contracts on negative test suites

Explanation

Correct Answer

A:

The accepted change is to resolve deployment and endpoint from runtime configuration; this evaluation checks whether it can avoid routing production questions to a test model through hardcoded SDK settings, since endpoint and deployment identifiers are environment-specific; runtime configuration keeps inference calls aligned to the project.

Incorrect Answers

B:

This test concerns the ability to prevent malformed model-generated parameters from reaching a write-capable API, for which tool schemas guide argument shape, but the application must check types, constraints, identity and authorization before execution; it does not establish whether the change to resolve deployment and endpoint from runtime configuration works for the current release.

C:

This test concerns the ability to show partial assistant output to a user while stopping work after the request is canceled, for which streaming delivers incremental output; cancellation and backpressure prevent orphaned inference work and stale interface content; it does not establish whether the change to resolve deployment and endpoint from runtime configuration works for the current release.

D:

This test concerns the ability to verify that an app recovers safely from empty messages, long context, and invalid tool responses, for which boundary-case testing surfaces parsing and orchestration failures that standard benchmark prompts miss; it does not establish whether the change to resolve deployment and endpoint from runtime configuration works for the current release.

Question 175

A pilot has a functional gap: A prompt revision sounds better in a few demos but worsens several critical user tasks. How should the team address that gap?

  1. Evaluate claim support against the supplied passages and require repair or abstention on failures
  2. Add question-answer relevance evaluations with labeled examples and failure review
  3. Run A/B offline comparisons on a representative fixed dataset before rollout
  4. Detect empty or low-confidence retrieval and return a supported no-answer or escalation response

Explanation

Correct Answer

C:

The fault signal is that a prompt revision sounds better in a few demos but worsens several critical user tasks; the targeted fix works because paired evaluation on identical cases isolates prompt impact; production traffic shifts make uncontrolled comparisons misleading, restoring the ability to choose between candidate prompt versions by measured task quality rather than intuition.

Incorrect Answers

A:

If the failing function were to catch plausible statements that are not substantiated by the retrieved evidence, this change could be relevant because faithfulness checks whether generated claims follow provided source material, not whether wording is eloquent; the actual symptom is that a prompt revision sounds better in a few demos but worsens several critical user tasks, which calls for a different repair.

B:

If the failing function were to detect answers that are grounded in retrieved text but do not address the actual question, this change could be relevant because relevance evaluates how directly a response serves the request; factual grounding and response usefulness are related but distinct axes; the actual symptom is that a prompt revision sounds better in a few demos but worsens several critical user tasks, which calls for a different repair.

D:

If the failing function were to avoid fabricating an answer to a question not covered by the organization's knowledge collection, this change could be relevant because an evidence threshold and abstention path distinguish unknown facts from model completion skill; confident text is not proof of source support; the actual symptom is that a prompt revision sounds better in a few demos but worsens several critical user tasks, which calls for a different repair.

Question 176

After deployment, the team observes: A legal-summary app produces a fluent conclusion absent from the document. Which remediation addresses the likely cause?

  1. Request structured outputs conforming to the expected schema and validate before ingestion
  2. Add malformed-input fixtures and explicit parser error handling to the integration tests
  3. Evaluate claim support against the supplied passages and require repair or abstention on failures
  4. Attach retrieved chunk references to the claims they substantiate and validate that mapping
  5. Add question-answer relevance evaluations with labeled examples and failure review

Explanation

Correct Answer

C:

The fault signal is that a legal-summary app produces a fluent conclusion absent from the document; the targeted fix works because faithfulness checks whether generated claims follow provided source material, not whether wording is eloquent, restoring the ability to catch plausible statements that are not substantiated by the retrieved evidence.

Incorrect Answers

A:

If the failing function were to send model-produced classifications into a strict downstream API contract, this change could be relevant because schema-constrained generation helps enforce machine-readable fields and types; application-level validation is still needed for business rules; the actual symptom is that a legal-summary app produces a fluent conclusion absent from the document, which calls for a different repair.

B:

If the failing function were to verify that an app recovers safely from empty messages, long context, and invalid tool responses, this change could be relevant because boundary-case testing surfaces parsing and orchestration failures that standard benchmark prompts miss; the actual symptom is that a legal-summary app produces a fluent conclusion absent from the document, which calls for a different repair.

D:

If the failing function were to let a compliance reviewer locate the specific evidence supporting each generated assertion, this change could be relevant because citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; the actual symptom is that a legal-summary app produces a fluent conclusion absent from the document, which calls for a different repair.

E:

If the failing function were to detect answers that are grounded in retrieved text but do not address the actual question, this change could be relevant because relevance evaluates how directly a response serves the request; factual grounding and response usefulness are related but distinct axes; the actual symptom is that a legal-summary app produces a fluent conclusion absent from the document, which calls for a different repair.

Question 177

A design review proposes to implement fallback when retrieval is insufficient. Which measurable check would establish success on real inputs?

  1. Inspect request-target metadata and run an environment-isolation smoke test
  2. Check claim-to-citation accuracy on questions with several similar policy sections
  3. Test for cross-department leakage and expired-document citation rate
  4. Measure unsupported-answer rate and correct abstentions on out-of-scope questions

Explanation

Correct Answer

D:

The accepted change is to implement fallback when retrieval is insufficient; this evaluation checks whether it can avoid fabricating an answer to a question not covered by the organization's knowledge collection, since an evidence threshold and abstention path distinguish unknown facts from model completion skill; confident text is not proof of source support.

Incorrect Answers

A:

This test concerns the ability to avoid routing production questions to a test model through hardcoded SDK settings, for which endpoint and deployment identifiers are environment-specific; runtime configuration keeps inference calls aligned to the project; it does not establish whether the change to implement fallback when retrieval is insufficient works for the current release.

B:

This test concerns the ability to let a compliance reviewer locate the specific evidence supporting each generated assertion, for which citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; it does not establish whether the change to implement fallback when retrieval is insufficient works for the current release.

C:

This test concerns the ability to ensure answers include only documents for a user's department and effective policy year, for which retrieval filters narrow the evidence set by structured attributes before the model sees it; prompt instructions are not security filters; it does not establish whether the change to implement fallback when retrieval is insufficient works for the current release.

Question 178

The team plans to measure factual faithfulness to source context. Which test provides the clearest evidence that this change works?

  1. Track relevance scores separately from citation-groundedness metrics
  2. Trace required step ordering and block completions that omit a mandatory stage
  3. Compute unsupported-claim rate on source-grounded summaries
  4. Count handled failures and confirm consistent error contracts on negative test suites

Explanation

Correct Answer

C:

The accepted change is to measure factual faithfulness to source context; this evaluation checks whether it can catch plausible statements that are not substantiated by the retrieved evidence, since faithfulness checks whether generated claims follow provided source material, not whether wording is eloquent.

Incorrect Answers

A:

This test concerns the ability to detect answers that are grounded in retrieved text but do not address the actual question, for which relevance evaluates how directly a response serves the request; factual grounding and response usefulness are related but distinct axes; it does not establish whether the change to measure factual faithfulness to source context works for the current release.

B:

This test concerns the ability to execute a known approval workflow where inputs must be validated before a model drafts a response, for which explicit workflow stages make ordering and failure handling reproducible; free-form agent planning is unnecessary for a fixed process; it does not establish whether the change to measure factual faithfulness to source context works for the current release.

D:

This test concerns the ability to verify that an app recovers safely from empty messages, long context, and invalid tool responses, for which boundary-case testing surfaces parsing and orchestration failures that standard benchmark prompts miss; it does not establish whether the change to measure factual faithfulness to source context works for the current release.

Question 179

Production telemetry reveals this failure: An agent waits sequentially for three unrelated inventory-read APIs, making responses unnecessarily slow. Which change most directly corrects it?

  1. Use deterministic status rules for routing and reserve the LLM for language-specific subtasks
  2. Attach retrieved chunk references to the claims they substantiate and validate that mapping
  3. Execute independent authorized reads in parallel and join results before generation
  4. Implement a controlled request-tool-result loop with a maximum number of steps

Explanation

Correct Answer

C:

The fault signal is that an agent waits sequentially for three unrelated inventory-read APIs, making responses unnecessarily slow; the targeted fix works because parallelizing independent operations shortens critical-path latency but is unsafe when one action requires another's output, restoring the ability to reduce latency when separate read-only lookups have no ordering dependency.

Incorrect Answers

A:

If the failing function were to choose among distinct business flows based on a validated account status, this change could be relevant because rules engines offer predictable decisions for simple structured conditions, while language models handle unstructured interpretation; the actual symptom is that an agent waits sequentially for three unrelated inventory-read APIs, making responses unnecessarily slow, which calls for a different repair.

B:

If the failing function were to let a compliance reviewer locate the specific evidence supporting each generated assertion, this change could be relevant because citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; the actual symptom is that an agent waits sequentially for three unrelated inventory-read APIs, making responses unnecessarily slow, which calls for a different repair.

D:

If the failing function were to allow a model to request multiple independent data lookups while limiting cost and duration, this change could be relevant because an orchestration loop can inspect tool requests, execute authorized tools, return results, and terminate under explicit limits; the actual symptom is that an agent waits sequentially for three unrelated inventory-read APIs, making responses unnecessarily slow, which calls for a different repair.

Question 180

A product team needs its AI workflow to let a compliance reviewer locate the specific evidence supporting each generated assertion. What technical approach most directly enables that?

  1. Use passage-level citations in RAG answers
  2. Use constrained structured JSON output
  3. Implement fallback when retrieval is insufficient
  4. Use chunk overlap for boundary-spanning facts

Explanation

Correct Answer

A:

For the stated requirement to let a compliance reviewer locate the specific evidence supporting each generated assertion, citations tied to retrieved chunks improve traceability; a source list without claim alignment does not prove each claim; this is the mechanism that the system must implement.

Incorrect Answers

B:

For a system whose objective is to send model-produced classifications into a strict downstream API contract, schema-constrained generation helps enforce machine-readable fields and types; application-level validation is still needed for business rules; this does not directly meet the required outcome to let a compliance reviewer locate the specific evidence supporting each generated assertion.

C:

For a system whose objective is to avoid fabricating an answer to a question not covered by the organization's knowledge collection, an evidence threshold and abstention path distinguish unknown facts from model completion skill; confident text is not proof of source support; this does not directly meet the required outcome to let a compliance reviewer locate the specific evidence supporting each generated assertion.

D:

For a system whose objective is to keep context that crosses page or section boundaries from being lost during indexing, chunk overlap can retain nearby context across splits, although excessive overlap inflates index size and duplicates retrieval; this does not directly meet the required outcome to let a compliance reviewer locate the specific evidence supporting each generated assertion.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests