Important Anthropic CCAR-F Exam Questions

Answer: A

Explanation:

Option A implements an iterative orchestrator-worker loop in which synthesis is treated as an evaluation checkpoint rather than an irreversible transition to report generation. When synthesis identifies missing evidence, the coordinator can formulate focused follow-up tasks, send them to the agents with the appropriate tools, and repeat synthesis after receiving the additional findings.

Anthropic's multi-agent research architecture follows this pattern: the lead agent synthesizes subagent findings and determines whether further research is required. If gaps remain, it creates additional subagents or refines the research strategy. This preserves specialization and centralized control.

Option B makes the incompleteness visible but does not improve research coverage. Option C may increase initial cost without guaranteeing that the specific gaps discovered during synthesis will be addressed. Option D weakens separation of concerns by giving the synthesis agent search capabilities that belong to the research specialists. Targeted redelegation is more efficient because the second research round is informed by concrete deficiencies rather than speculative breadth. It also allows the coordinator to track completeness explicitly before authorizing final report generation.

================

Answer: A

Explanation:

An aggregate accuracy value can conceal severe performance disparities. A system may achieve 97% overall accuracy while performing poorly on a low-volume document type, a critical financial field, or a specific edge case. Automating outputs solely from the aggregate figure could therefore expose downstream systems to concentrated, high-impact errors.

Anthropic's evaluation guidance states that evaluations should be task-specific, reflect the real-world task distribution, and explicitly include edge cases. It also emphasizes multidimensional success criteria rather than reliance on a single global metric. (https://docs.anthropic.com/en/docs/build-with-claude/develop-tests) Option A applies those principles by stratifying performance according to document type and field. This reveals whether confidence is calibrated consistently and whether the proposed automation threshold remains safe for every operationally significant segment.

Option B is useful only after segment-level performance has been understood. Selecting a global threshold cannot correct a subgroup where confidence is systematically overstated. Option C is necessary governance work, but it treats the overall 97% result as though errors were uniformly distributed. Option D places unvalidated outputs into downstream systems and depends on passive error reporting, which may fail to detect silent corruption.

The correct deployment gate is therefore segmented validation, followed by threshold selection, downstream acceptance criteria, and a controlled pilot.

Official references/topics: Define Success Criteria; Task-Specific Evaluations; Edge-Case Coverage; Reliability Segmentation.

Answer: C

Explanation:

The decisive trade-off is whether the workflow can tolerate asynchronous completion. Anthropic's Message Batches API documentation states that batch requests receive a 50% discount and that most batches finish within one hour, but processing can continue for as long as 24 hours. Therefore, the review process is suitable for batching only if potentially delayed feedback remains useful.

Option C directly tests that operational requirement. The reviews are already non-blocking, which makes batching promising, but non-blocking does not automatically mean that feedback delivered many hours later still has value. If developers have already merged and moved to unrelated work, delayed security findings may increase remediation cost or remain unaddressed.

Option A is manageable through each request's unique custom_id, which allows results to be correlated regardless of return order. Option B is incorrect because batch requests can contain supported multi-turn conversation histories and tool-use configurations; each request is processed independently, but it need not represent a single conversational turn. Option D reverses the relevant consideration: batch processing is not intended to provide near-instantaneous responses. The real question is whether the cost saving justifies its longer and nondeterministic completion time.

================

Answer: A

Explanation:

Option A is correct because the failure results from excessive output volume, not from JSON or tool use. Anthropic documents that max_tokens is a hard output ceiling and that a response stopped at this limit reports stop_reason: 'max_tokens'. Its tool-streaming guidance also warns that generation can stop midway through a tool parameter, leaving partial JSON that must not be treated as complete. Splitting the pull request into bounded file groups limits the number of findings produced by each call, preserves the report_findings schema, and allows the pipeline to validate every response independently before merging and deduplicating the arrays. Shared dependency context can still be included when cross-file analysis is required.

Option B only postpones the failure because the model's output maximum remains finite. Option C abandons the structured contract and makes deterministic parsing less reliable. Option D changes the review requirements by suppressing potentially valuable lower-severity findings, while the retried request could still reach the same limit. Controlled partitioning followed by deterministic aggregation is therefore the most reliable and scalable design. Anthropic stop-reason guidance and tool-streaming guidance.

Answer: A

Explanation:

Option A is the most sustainable open-world design. The property-type domain is demonstrably not closed: uncommon values already occur, and new labels continue to appear. An enum is appropriate only when the valid vocabulary is stable and exhaustive. If the source contains a legitimate value outside that vocabulary, the schema either rejects it or pressures the model to map it inaccurately. A free-form string preserves the source meaning, while deterministic post-processing can normalize ''studio,'' ''loft,'' ''duplex,'' and future terms into an evolving canonical taxonomy. Anthropic's Structured Outputs documentation distinguishes structural schema conformance from the semantic decisions made within that schema; it also documents that JSON Schema constraints such as enums govern allowable output shapes and values. Few-shot mapping to the nearest existing enum, option B, discards real distinctions. Continually expanding the enum, option C, creates recurring schema changes and still lags newly observed values. An ''other'' bucket with detail, option D, is workable but collapses all emerging categories at the primary field and complicates analytics until downstream interpretation occurs. Capturing the raw type first and normalizing it in code provides lossless extraction, testable mappings, and forward compatibility.

================