Anthropic CCAR-P (Claude Certified Architect - Professional) Practice Test 71

Question 700

A media licensing pilot has enabled Hybrid lexical and semantic retrieval. Because semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases, the release board wants evidence of effectiveness, not just a favorable demonstration. Which metric most directly tests the proposed control?

  1. Measure eligible-document retrieval precision after filtering
  2. Measure cross-boundary answer recall versus indexed token volume
  3. Measure time-to-index-freshness and stale-answer incidence
  4. Measure recall on exact-code and paraphrase query subsets

Explanation

Correct Answer

D:

An appropriate gate measures recall on exact-code and paraphrase query subsets. It tests the mechanism in which the integration can combine keyword matches with embeddings and rank the merged results, rather than a cosmetic demonstration. That distinction matters because semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

Incorrect Answers

A:

Where questions require filtering by jurisdiction and policy effective date, a relevant check would establish whether the integration can store reliable source metadata for filtering before similarity search; but the acceptance gate must instead inspect recall on exact-code and paraphrase query subsets to detect the targeted failure.

B:

Where factual statements often cross page or paragraph cut points, a relevant check would establish whether the integration can use measured overlap where splitting otherwise loses neighboring facts; by contrast, the media licensing release needs evidence from recall on exact-code and paraphrase query subsets.

C:

When source policies change daily, retired versions must stop appearing and the integration can synchronize create-update-delete events into the retrieval index; however, that proxy would not demonstrate recall on exact-code and paraphrase query subsets for the selected intervention.

Question 701

In an insurance processing modernization project, drafted release notes repeatedly miss mandatory restrictions. Which approach best fits this specific operating constraint?

  1. Introduce Sequential dependency-aware handoffs
  2. Introduce Centralized permission broker for agents
  3. Introduce Supervisor-mediated delegation
  4. Introduce Independent critic with bounded revisions
  5. Introduce Parallel independent subtask fan-out

Explanation

Correct Answer

D:

The architectural distinction is that a separate evaluation pass checks defined criteria and requests capped revisions. The decision should be implemented with an independent rubric and revision limit and evaluated against rubric pass rate after a fixed maximum revision count; otherwise, unbounded self-critique loops can raise cost without a convergence rule.

Incorrect Answers

A:

When a reviewer cannot assess a draft before verified citations are available, the architecture can orchestrate a staged dependency graph with explicit prerequisites; whereas the insurance processing case must guard against the outcome that unbounded self-critique loops can raise cost without a convergence rule.

B:

When specialist agents request privileged data operations during delegated work, a service-side broker verifies user authority for each proposed action; the missing operational behavior is specific: a separate evaluation pass checks defined criteria and requests capped revisions.

C:

For a workload where several specialists must cooperate without granting every one all tools, a supervisor assigns tasks and consolidates results within distinct permission scopes; the current decision instead depends on a specific capability: a separate evaluation pass checks defined criteria and requests capped revisions.

E:

For a workload where multiple independent document checks dominate total processing time, the architecture can run independent checks concurrently and combine results with deterministic joins; the missing operational behavior is specific: a separate evaluation pass checks defined criteria and requests capped revisions.

Question 702

An independent assessment of an energy operations Claude system reports that semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases. Which change best closes the exposed control gap?

  1. Apply Hybrid lexical and semantic retrieval
  2. Apply Overlapping chunk strategy
  3. Apply Parent-child document retrieval
  4. Apply Metadata-rich index entries

Explanation

Correct Answer

A:

The reported failure calls for the control under which the integration can combine keyword matches with embeddings and rank the merged results. A prompt-only reminder or unrelated metric would not address the risk that semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases. Evidence of resolution requires synchronized lexical and vector indexes and an improvement in recall on exact-code and paraphrase query subsets.

Incorrect Answers

B:

This addresses the failure where excessive overlap inflates storage and duplicates retrieved context, because the integration can use measured overlap where splitting otherwise loses neighboring facts; yet it would not correct the observed weakness: semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

C:

Although the alternative reduces the risk that returning only tiny isolated fragments removes governing context because the integration can search fine-grained chunks then return appropriate parent spans for grounding; but this remedy operates on another boundary, not the case where semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

D:

When vector proximity cannot enforce policy version or residency constraints, the integration can store reliable source metadata for filtering before similarity search; whereas the energy operations defect follows the path in which semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

Question 703

A healthcare operations service owner is reviewing Unsupported-claim provenance analysis after a canary rollout. The critical regression would occur if adding more citation markers can disguise hallucinations. What observable acceptance signal belongs in the dashboard?

  1. Audit held-out multi-hop reasoning gain from model change
  2. Audit unsupported claim rate by generation stage
  3. Audit retrieval recall at k on missing-rule cases
  4. Audit generalization gap across paraphrase and novelty subsets

Explanation

Correct Answer

B:

The correct signal is unsupported claim rate by generation stage; it must be computed for representative workloads and compared with a baseline. It checks whether the evaluation process can validate claim-to-span entailment rather than just link presence while surfacing the exposure that adding more citation markers can disguise hallucinations.

Incorrect Answers

A:

Where complex cross-document inferences fail despite accurate retrieved passages, a relevant check would establish whether the evaluation process can evaluate a more capable tier on the same evidence and instructions; but the acceptance gate must instead inspect unsupported claim rate by generation stage to detect the targeted failure.

C:

When generated answers omit a rule even though it exists in the knowledge base, the evaluation process can inspect whether relevant source passages reached the model. A favorable result on that test would not establish unsupported claim rate by generation stage for this workflow, where a response confidently cites text that does not support the claim; however, that proxy would not demonstrate unsupported claim rate by generation stage for the selected intervention.

D:

The alternative indicator checks whether the evaluation process can test whether examples bias the model to surface patterns rather than policy rules when a few-shot classifier works on familiar categories but fails novel wording; but the acceptance gate must instead inspect unsupported claim rate by generation stage to detect the targeted failure.

Question 704

A public-sector casework pilot cannot be accepted because a single SLA overcommits low-priority work or underprotects critical flows. The owner can change one primary architectural control; what should it be?

  1. Prioritize Service-tier differentiation
  2. Prioritize Business outcome SLO
  3. Prioritize Structured feedback intake
  4. Prioritize Change-communication plan

Explanation

Correct Answer

A:

The remediation acts at the relevant design boundary: the delivery team can commit separate targets and fallback behaviors by priority class. That counteracts the observed hazard that a single SLA overcommits low-priority work or underprotects critical flows. Rollout should include workload classification and capacity allocation, with regression evidence from sLA compliance by request tier under load.

Incorrect Answers

B:

When HTTP availability alone does not measure decision quality, the delivery team can specify user-observable success such as validated resolutions and response time; whereas the public-sector casework defect follows the path in which a single SLA overcommits low-priority work or underprotects critical flows.

C:

Although the alternative reduces the risk that anecdotal messages without context are hard to triage because the delivery team can capture task type, evidence, severity and version in one feedback queue; yet it would not correct the observed weakness: a single SLA overcommits low-priority work or underprotects critical flows.

D:

This addresses the failure where silent behavior changes can violate expectations and training materials, because the delivery team can announce consequential changes, effective dates and rollback contacts; but this remedy operates on another boundary, not the case where a single SLA overcommits low-priority work or underprotects critical flows.

Question 705

The post-incident analysis for telecommunications support concludes that no-review automation can silently degrade after a distribution shift. Which remediation targets the relevant failure boundary?

  1. Introduce Escalation timeout and fail-closed route
  2. Introduce Risk-tiered approval workflow
  3. Introduce Reviewer evidence packet
  4. Introduce Override reason capture
  5. Introduce Random audit of low-risk automation

Explanation

Correct Answer

E:

This intervention addresses the failure at its source because the control layer can sample approved outputs for independent periodic quality review. To keep the issue from recurring, establish stratified audit plan and escalation thresholds and verify detected error rate and audit sampling coverage; otherwise no-review automation can silently degrade after a distribution shift can persist.

Incorrect Answers

A:

This addresses the failure where automatic execution after a timeout defeats the approval requirement, because the control layer can hold or safely reject high-impact actions until review is complete; whereas the telecommunications support defect follows the path in which no-review automation can silently degrade after a distribution shift.

B:

When blanket approval for every step wastes capacity; blanket autonomy hides risk, the control layer can send consequential actions to authorized reviewers while allowing bounded automation elsewhere; yet it would not correct the observed weakness: no-review automation can silently degrade after a distribution shift.

C:

Although the alternative reduces the risk that an approval button without context cannot support an accountable decision because the control layer can present retrieved evidence, uncertainty and proposed effect alongside each action; but this remedy operates on another boundary, not the case where no-review automation can silently degrade after a distribution shift.

D:

Where reviewers often correct suggestions but no explanation is recorded, the control layer can capture structured reasons and attach them to evaluation feedback; whereas the telecommunications support defect follows the path in which no-review automation can silently degrade after a distribution shift.

Question 706

A research administration release candidate fails review after assessors observe that free-form numeric reasoning can introduce avoidable calculation mistakes. What intervention should take priority?

  1. Introduce Decomposed multi-pass reasoning
  2. Introduce Structured intermediate checklist
  3. Introduce Deterministic external computation
  4. Introduce Zero-shot baseline

Explanation

Correct Answer

C:

The engineering fix changes the mechanism rather than the presentation: the model configuration can use code or a trusted service for calculations and Claude for interpretation. It requires authorized calculation tool and input validation; a regression test should examine exact numeric pass rate across boundary values in cases where free-form numeric reasoning can introduce avoidable calculation mistakes.

Incorrect Answers

A:

Where a negotiation report contains independent arithmetic and contractual judgments, the model configuration can compute verifiable components separately before final synthesis; yet it would not correct the observed weakness: free-form numeric reasoning can introduce avoidable calculation mistakes.

B:

When a free-form summary can omit a required category even when fluent, the model configuration can require a concise explicit checklist of required criteria before final output; but this remedy operates on another boundary, not the case where free-form numeric reasoning can introduce avoidable calculation mistakes.

D:

When examples can add maintenance cost without measurable gain, the model configuration can start with concise instructions before adding demonstrations that consume context; yet it would not correct the observed weakness: free-form numeric reasoning can introduce avoidable calculation mistakes.

Question 707

A procurement services change board is considering removing Bounded issue-to-patch automation to simplify a release. Which outcome is the primary residual risk if that protection is absent?

  1. The danger that unattributed code changes weaken auditability and incident diagnosis
  2. The danger that one general review prompt may miss specialist concerns
  3. The danger that a single sweeping rewrite is hard to validate and roll back
  4. The danger that end-to-end unattended merging removes an independent quality gate

Explanation

Correct Answer

D:

The relevant failure mode is that end-to-end unattended merging removes an independent quality gate. The existing design can prevent it when the engineering workflow can allow an agent to propose a patch and tests within a restricted repository scope. The team needs issue reproduction, scoped permissions and merge protection and must monitor accepted patches without policy or test regressions to support any claim of equivalent protection.

Incorrect Answers

A:

When release teams cannot tell why generated modifications were proposed, unattributed code changes weaken auditability and incident diagnosis; but this is not the immediate loss associated with omitting Bounded issue-to-patch automation: end-to-end unattended merging removes an independent quality gate.

B:

When multiple quality concerns need independent assessment of a large change, one general review prompt may miss specialist concerns; however, the named safeguard addresses the distinct hazard that end-to-end unattended merging removes an independent quality gate.

C:

When a large codebase migration would touch many coupled modules, a single sweeping rewrite is hard to validate and roll back; whereas removing Bounded issue-to-patch automation primarily exposes the risk that end-to-end unattended merging removes an independent quality gate.

Question 708

The contract administration architect proposes omitting Hybrid lexical and semantic retrieval from the production design. What is the most direct failure mode that decision would expose?

  1. A failure where returning only tiny isolated fragments removes governing context
  2. A failure where semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases
  3. A failure where fixed arbitrary cuts can detach qualifiers from the associated rule
  4. A failure where stale embeddings can support obsolete guidance after source correction

Explanation

Correct Answer

B:

The risk is not a generic model inconvenience: semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases. The proposed safeguard changes that boundary because the integration can combine keyword matches with embeddings and rank the merged results, and the release would need synchronized lexical and vector indexes plus recall on exact-code and paraphrase query subsets to show the risk remains contained.

Incorrect Answers

A:

The competing hazard is that returning only tiny isolated fragments removes governing context in a workload where short chunks are searchable but answers need full contractual sections; however, the named safeguard addresses the distinct hazard that semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

C:

The competing hazard is that fixed arbitrary cuts can detach qualifiers from the associated rule in a workload where policies contain headings and exceptions that must stay together; but this is not the immediate loss associated with omitting Hybrid lexical and semantic retrieval: semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

D:

If source policies change daily and retired versions must stop appearing, stale embeddings can support obsolete guidance after source correction; however, the named safeguard addresses the distinct hazard that semantic-only retrieval can miss exact identifiers while lexical-only misses paraphrases.

Question 709

The contract administration go-live board accepts the premise that a separate evaluation pass checks defined criteria and requests capped revisions. Which preparation is indispensable before it can rely on Independent critic with bounded revisions operationally?

  1. Prepare structured handoff payloads and stage readiness checks
  2. Prepare monitored per-task budgets and fail-safe exit messages
  3. Prepare role-specific tool grants and aggregation rules
  4. Prepare an independent rubric and revision limit
  5. Prepare user-context propagation and centralized authorization checks
  6. Prepare versioned data contracts and ownership of identifiers

Explanation

Correct Answer

D:

What makes the control operable is an independent rubric and revision limit. It supplies the precondition for the mechanism whereby a separate evaluation pass checks defined criteria and requests capped revisions. The residual failure to guard against is that unbounded self-critique loops can raise cost without a convergence rule, measured through rubric pass rate after a fixed maximum revision count.

Incorrect Answers

A:

The setup is useful to enable the alternative's behavior where the architecture can orchestrate a staged dependency graph with explicit prerequisites and a reviewer cannot assess a draft before verified citations are available; however, the named intervention needs an independent rubric and revision limit before its protection can be enforced.

B:

Where a planner sometimes delegates recursively until requests time out, the architecture can impose step, token, cost and time ceilings with explicit exit behavior; but the production dependency for Independent critic with bounded revisions is an independent rubric and revision limit.

C:

The setup is useful to enable the alternative's behavior where a supervisor assigns tasks and consolidates results within distinct permission scopes and several specialists must cooperate without granting every one all tools; for this release, the team must first verify an independent rubric and revision limit.

E:

For the separate challenge where specialist agents request privileged data operations during delegated work, a service-side broker verifies user authority for each proposed action; but the production dependency for Independent critic with bounded revisions is an independent rubric and revision limit.

F:

For the separate challenge where teams merge agent outputs that refer to inconsistent record identifiers, common schemas and immutable artifact versions make specialist outputs composable; for this release, the team must first verify an independent rubric and revision limit.

Question 710

Following a regression in healthcare operations, the team establishes that adding more citation markers can disguise hallucinations. Which proposed redesign most directly prevents recurrence?

  1. Introduce Prompt ambiguity isolation
  2. Introduce Overfit-to-examples analysis
  3. Introduce Context truncation investigation
  4. Introduce Unsupported-claim provenance analysis
  5. Introduce Retrieval recall diagnosis

Explanation

Correct Answer

D:

The remediation acts at the relevant design boundary: the evaluation process can validate claim-to-span entailment rather than just link presence. That counteracts the observed hazard that adding more citation markers can disguise hallucinations. Rollout should include source spans and expert claim verification, with regression evidence from unsupported claim rate by generation stage.

Incorrect Answers

A:

Although the alternative reduces the risk that changing all components at once prevents root-cause attribution because the evaluation process can test instruction clarity while holding model and retrieval constant; but this remedy operates on another boundary, not the case where adding more citation markers can disguise hallucinations.

B:

The alternative reduces the risk that more examples of the same type can worsen coverage of unseen cases because the evaluation process can test whether examples bias the model to surface patterns rather than policy rules; however, the healthcare operations defect follows the path in which adding more citation markers can disguise hallucinations.

C:

When increasing temperature cannot restore clipped evidence, the evaluation process can inspect token budgets and the presence of decisive passages after packing; yet it would not correct the observed weakness: adding more citation markers can disguise hallucinations.

E:

When rewriting prompts cannot compensate for evidence never retrieved, the evaluation process can inspect whether relevant source passages reached the model; whereas the healthcare operations defect follows the path in which adding more citation markers can disguise hallucinations.

Claude Certified Architect - Professional (CCAR-P) Practice Tests