Question 510
An incident postmortem for a policy document normalization service requires controls so that invalid parameters lead to a clear correction rather than a blind retry and that a model learns a boundary between superficially similar cases. Which paired remedy covers both failure modes?
- hide the invalid field behind a generic internal-error message; and show only easy positive samples
- hide the invalid field behind a generic internal-error message; and include contrasting positive and negative demonstrations
- return a non-retryable validation error with the field and expected constraint; and show only easy positive samples
- return a non-retryable validation error with the field and expected constraint; and include contrasting positive and negative demonstrations
Explanation
Correct Answer
D:In a policy document normalization service, field-specific feedback lets the caller fix the input before resubmission; alongside this, near-boundary examples communicate the decisive distinction and reduce false positives, satisfying the distinct requirements that invalid parameters lead to a clear correction rather than a blind retry and that a model learns a boundary between superficially similar cases.
Incorrect Answers
A:In a policy document normalization service, the client lacks the information needed to repair the request and the model may generalize the label to unrelated but similar-looking items; the design would fail both conditions: invalid parameters lead to a clear correction rather than a blind retry and a model learns a boundary between superficially similar cases.
B:In a policy document normalization service, near-boundary examples communicate the decisive distinction and reduce false positives, but the client lacks the information needed to repair the request; therefore the separate requirement that invalid parameters lead to a clear correction rather than a blind retry would remain unsatisfied.
C:In a policy document normalization service, field-specific feedback lets the caller fix the input before resubmission, but the model may generalize the label to unrelated but similar-looking items; therefore the separate requirement that a model learns a boundary between superficially similar cases would remain unsatisfied.
Question 511
A team operating a structured report conversion service wants to establish two separate behaviors: a classification uses a limited approved category set; in addition, persistent ambiguous data receives human review. Which plan implements both?
- define precise output labels and mapping rules; and stop after defined retries and route unresolved cases to a reviewer
- define precise output labels and mapping rules; and keep retrying until the model finally emits something that passes
- allow a free-form category for each case; and stop after defined retries and route unresolved cases to a reviewer
- allow a free-form category for each case; and keep retrying until the model finally emits something that passes
Explanation
Correct Answer
A:In a structured report conversion service, controlled categories reduce schema drift and make downstream analytics comparable; alongside this, a bounded escalation path prevents infinite loops and invented resolutions, satisfying the distinct requirements that a classification uses a limited approved category set and that persistent ambiguous data receives human review.
Incorrect Answers
B:In a structured report conversion service, controlled categories reduce schema drift and make downstream analytics comparable, but unlimited retries waste resources and can reinforce guesses; therefore the separate requirement that persistent ambiguous data receives human review would remain unsatisfied.
C:In a structured report conversion service, a bounded escalation path prevents infinite loops and invented resolutions, but unbounded labels fragment the intended taxonomy; therefore the separate requirement that a classification uses a limited approved category set would remain unsatisfied.
D:In a structured report conversion service, unbounded labels fragment the intended taxonomy and unlimited retries waste resources and can reinforce guesses; the design would fail both conditions: a classification uses a limited approved category set and persistent ambiguous data receives human review.
Question 512
Tests for a purchase-order parsing service will pass only if a synthesized statement can be traced to its supporting source and if a conclusion supported by model agreement is also tested against source evidence. Which proposed design directly satisfies both checks?
- publish a source list detached from the statements it supports; and require source and business-rule checks regardless of model agreement
- publish a source list detached from the statements it supports; and promote agreement alone to a guaranteed correctness signal
- attach source IDs and relevant excerpts to individual claims; and promote agreement alone to a guaranteed correctness signal
- attach source IDs and relevant excerpts to individual claims; and require source and business-rule checks regardless of model agreement
Explanation
Correct Answer
D:In a purchase-order parsing service, claim-level provenance enables an auditor to verify each assertion; alongside this, correlated model errors can survive consensus without external verification, satisfying the distinct requirements that a synthesized statement can be traced to its supporting source and that a conclusion supported by model agreement is also tested against source evidence.
Incorrect Answers
A:In a purchase-order parsing service, correlated model errors can survive consensus without external verification, but a bibliography alone does not show which evidence backs each claim; therefore the separate requirement that a synthesized statement can be traced to its supporting source would remain unsatisfied.
B:In a purchase-order parsing service, a bibliography alone does not show which evidence backs each claim and shared biases can produce identical wrong conclusions; the design would fail both conditions: a synthesized statement can be traced to its supporting source and a conclusion supported by model agreement is also tested against source evidence.
C:In a purchase-order parsing service, claim-level provenance enables an auditor to verify each assertion, but shared biases can produce identical wrong conclusions; therefore the separate requirement that a conclusion supported by model agreement is also tested against source evidence would remain unsatisfied.
Question 513
For a monorepo testing overhaul, acceptance depends on demonstrating that the review produces a small, auditable patch. Which approach is the strongest fit?
- turn the feedback into a concrete test or change request
- inspect failure evidence and revise the smallest relevant change
- replace every file in the component to ensure consistency
- iterate with focused diffs rather than wholesale regeneration
Explanation
Correct Answer
D:For a monorepo testing overhaul, narrow changes are easier to test and evaluate against explicit requirements; this directly establishes that the review produces a small, auditable patch instead of relying on an unrelated workflow control.
Incorrect Answers
A:Although actionable feedback narrows the next edit and makes improvement measurable, that only supports the distinct goal that reviewer feedback points to a specific missed requirement, not the requirement that the review produces a small, auditable patch in a monorepo testing overhaul.
B:Although test-driven feedback pinpoints a specific unmet behavior for the next iteration, that only supports the distinct goal that a patch that fails regression tests is diagnosed and corrected, not the requirement that the review produces a small, auditable patch in a monorepo testing overhaul.
C:In a monorepo testing overhaul, unnecessary regeneration can erase working behavior; as a result, the stated requirement that the review produces a small, auditable patch would still not be met.
Question 514
In a technical review of a CLI application maintenance task, the architect must address the following constraints together: multiple subagent outputs can be traced to their individual tasks; a new agent gets only durable context from a completed job. Which change set is most defensible?
- attach task IDs and source references to returned artifacts; and paste a full historical transcript into every new session
- attach task IDs and source references to returned artifacts; and initialize a fresh session with a verified structured summary
- strip origin metadata from each response before merging; and paste a full historical transcript into every new session
- implement only initialize a fresh session with a verified structured summary; make no corresponding change for the first requirement
- implement only attach task IDs and source references to returned artifacts; make no corresponding change for the second requirement
- strip origin metadata from each response before merging; and initialize a fresh session with a verified structured summary
Explanation
Correct Answer
B:In a CLI application maintenance task, traceable results let the coordinator identify which delegation produced each claim; alongside this, a bounded handoff transfers necessary facts without carrying discarded trial-and-error context, satisfying the distinct requirements that multiple subagent outputs can be traced to their individual tasks and that a new agent gets only durable context from a completed job.
Incorrect Answers
A:In a CLI application maintenance task, traceable results let the coordinator identify which delegation produced each claim, but irrelevant turns can consume context and revive outdated assumptions; therefore the separate requirement that a new agent gets only durable context from a completed job would remain unsatisfied.
C:In a CLI application maintenance task, source-less aggregation impairs diagnosis and fact checking and irrelevant turns can consume context and revive outdated assumptions; the design would fail both conditions: multiple subagent outputs can be traced to their individual tasks and a new agent gets only durable context from a completed job.
D:In a CLI application maintenance task, a bounded handoff transfers necessary facts without carrying discarded trial-and-error context, but the independent mechanism for the other criterion is absent: traceable results let the coordinator identify which delegation produced each claim; as written, the design does not ensure that multiple subagent outputs can be traced to their individual tasks.
E:In a CLI application maintenance task, traceable results let the coordinator identify which delegation produced each claim, but the independent mechanism for the other criterion is absent: a bounded handoff transfers necessary facts without carrying discarded trial-and-error context; as written, the design does not ensure that a new agent gets only durable context from a completed job.
F:In a CLI application maintenance task, a bounded handoff transfers necessary facts without carrying discarded trial-and-error context, but source-less aggregation impairs diagnosis and fact checking; therefore the separate requirement that multiple subagent outputs can be traced to their individual tasks would remain unsatisfied.
Question 515
The architecture of a multi-repository troubleshooting agent must meet these three criteria: independent stages of a complex job run efficiently; near-overlapping retrieval tools are selected by actual boundary; separate code modules can be inspected in parallel. Choose THREE measures, one for each requirement.
- document examples and conditions where each tool must not be used
- make the second specialist wait for the first even though their modules are independent
- rely exclusively on short tool names for routing
- divide work along independent outputs and run those stages in parallel
- spawn independent specialists in the same delegation stage
- split the job into an arbitrary fixed chain of agents
Explanation
Correct Answers
A:For a multi-repository troubleshooting agent, contrastive guidance helps tool choice when resource names look similar, which directly establishes that near-overlapping retrieval tools are selected by actual boundary; the separate requirement that separate code modules can be inspected in parallel is handled by a different action.
D:For a multi-repository troubleshooting agent, dependency-aware decomposition avoids serial work when branches do not consume one another, which directly establishes that independent stages of a complex job run efficiently; the separate requirement that near-overlapping retrieval tools are selected by actual boundary is handled by a different action.
E:For a multi-repository troubleshooting agent, concurrent subagents can examine non-overlapping modules before the coordinator merges their reports, which directly establishes that separate code modules can be inspected in parallel; the separate requirement that independent stages of a complex job run efficiently is handled by a different action.
Incorrect Answers
B:For a multi-repository troubleshooting agent, serialization wastes time without supplying a dependency, so this option cannot establish that separate code modules can be inspected in parallel; correctly addressing independent stages of a complex job run efficiently elsewhere does not repair that deficiency.
C:For a multi-repository troubleshooting agent, names alone may not capture the distinction in scope or authorization, so this option cannot establish that near-overlapping retrieval tools are selected by actual boundary; correctly addressing separate code modules can be inspected in parallel elsewhere does not repair that deficiency.
F:For a multi-repository troubleshooting agent, a rigid sequence creates waits even where no dependency exists, so this option cannot establish that independent stages of a complex job run efficiently; correctly addressing near-overlapping retrieval tools are selected by actual boundary elsewhere does not repair that deficiency.
Question 516
An audit of a shipping-exception helpdesk asks how to make sure that a synthesized statement can be traced to its supporting source. Which implementation should be selected?
- record source timestamps and refresh time-sensitive evidence
- attach source IDs and relevant excerpts to individual claims
- carry citation identifiers and retrieval context through synthesis
- publish a source list detached from the statements it supports
Explanation
Correct Answer
B:For a shipping-exception helpdesk, claim-level provenance enables an auditor to verify each assertion; this directly establishes that a synthesized statement can be traced to its supporting source instead of relying on an unrelated workflow control.
Incorrect Answers
A:Although freshness determines whether an otherwise credible fact applies now, that only supports the distinct goal that fast-changing information is checked against its publication date, not the requirement that a synthesized statement can be traced to its supporting source in a shipping-exception helpdesk.
C:Although provenance metadata must remain attached when intermediate summaries are merged, that only supports the distinct goal that source identity survives multi-agent research handoffs, not the requirement that a synthesized statement can be traced to its supporting source in a shipping-exception helpdesk.
D:In a shipping-exception helpdesk, a bibliography alone does not show which evidence backs each claim; as a result, the stated requirement that a synthesized statement can be traced to its supporting source would still not be met.
Question 517
Production readiness for a subscription support service depends on ensuring that a status field can contain only approved business states and that the retry process does not create duplicate records. Select the most appropriate combined approach.
- accept any model-generated status description; and assign stable source identifiers and deduplicate writes
- enforce the approved status enum with structured output and validate it in the consumer; and insert a fresh record on every model retry
- enforce the approved status enum with structured output and validate it in the consumer; and assign stable source identifiers and deduplicate writes
- accept any model-generated status description; and insert a fresh record on every model retry
Explanation
Correct Answer
C:In a subscription support service, a constrained schema expresses the allowed labels and downstream validation rejects out-of-contract values; alongside this, idempotent processing separates repeated model attempts from committed business mutations, satisfying the distinct requirements that a status field can contain only approved business states and that the retry process does not create duplicate records.
Incorrect Answers
A:In a subscription support service, idempotent processing separates repeated model attempts from committed business mutations, but free-form status phrases can violate a fixed business-state contract; therefore the separate requirement that a status field can contain only approved business states would remain unsatisfied.
B:In a subscription support service, a constrained schema expresses the allowed labels and downstream validation rejects out-of-contract values, but repeated extraction attempts could create duplicate downstream state; therefore the separate requirement that the retry process does not create duplicate records would remain unsatisfied.
D:In a subscription support service, free-form status phrases can violate a fixed business-state contract and repeated extraction attempts could create duplicate downstream state; the design would fail both conditions: a status field can contain only approved business states and the retry process does not create duplicate records.
Question 518
In a research PDF metadata pipeline, the engineering team must guarantee that a failed extraction is retried with specific corrective information. What is the most appropriate control?
- record validation failure codes and field paths per attempt
- feed field-level validation errors back in a bounded retry
- repeat the same vague prompt after each failure
- retain the original source, response version, and retry feedback
- stop after defined retries and route unresolved cases to a reviewer
Explanation
Correct Answer
B:For a research PDF metadata pipeline, targeted feedback identifies what the next attempt must correct; this directly establishes that a failed extraction is retried with specific corrective information instead of relying on an unrelated workflow control.
Incorrect Answers
A:Although granular observability reveals schema weaknesses and document patterns, that only supports the distinct goal that the team can diagnose which fields repeatedly fail extraction, not the requirement that a failed extraction is retried with specific corrective information in a research PDF metadata pipeline.
C:In a research PDF metadata pipeline, unchanged requests do not reveal which constraint was violated; as a result, the stated requirement that a failed extraction is retried with specific corrective information would still not be met.
D:Although traceable attempts allow auditors to distinguish correction from silent overwrite, that only supports the distinct goal that a corrected attempt preserves a record of prior validation findings, not the requirement that a failed extraction is retried with specific corrective information in a research PDF metadata pipeline.
E:Although a bounded escalation path prevents infinite loops and invented resolutions, that only supports the distinct goal that persistent ambiguous data receives human review, not the requirement that a failed extraction is retried with specific corrective information in a research PDF metadata pipeline.
Question 519
An architecture review of a supplier invoice extractor finds two missing guarantees: a tool result includes enough fields to support audit and follow-on actions; the agent can recover exact policy facts after compaction. What two-part change closes both gaps?
- return only a friendly summary with no source identifier; and treat the compressed summary as an immutable source of truth
- return only a friendly summary with no source identifier; and retrieve authoritative facts from a durable store when needed
- return structured identifiers, status, and provenance with the data; and retrieve authoritative facts from a durable store when needed
- return structured identifiers, status, and provenance with the data; and treat the compressed summary as an immutable source of truth
Explanation
Correct Answer
C:In a supplier invoice extractor, machine-readable outputs let downstream stages validate exactly what was retrieved; alongside this, re-reading source state restores precision that a condensed history cannot guarantee, satisfying the distinct requirements that a tool result includes enough fields to support audit and follow-on actions and that the agent can recover exact policy facts after compaction.
Incorrect Answers
A:In a supplier invoice extractor, a paraphrase hides the record needed for verification or follow-up and summaries may simplify caveats or omit an exception; the design would fail both conditions: a tool result includes enough fields to support audit and follow-on actions and the agent can recover exact policy facts after compaction.
B:In a supplier invoice extractor, re-reading source state restores precision that a condensed history cannot guarantee, but a paraphrase hides the record needed for verification or follow-up; therefore the separate requirement that a tool result includes enough fields to support audit and follow-on actions would remain unsatisfied.
D:In a supplier invoice extractor, machine-readable outputs let downstream stages validate exactly what was retrieved, but summaries may simplify caveats or omit an exception; therefore the separate requirement that the agent can recover exact policy facts after compaction would remain unsatisfied.
Question 520
The release checklist for a tender-response data extractor explicitly checks that a partially failed batch is repaired efficiently; it also checks that an uncertain high-impact decision is routed for review instead of automatic execution. What coordinated solution meets the checklist?
- rerun the entire completed batch whenever one record fails; and proceed whenever the assistant's prose sounds certain
- rerun the entire completed batch whenever one record fails; and set a clear evidence or confidence boundary and require review below it
- identify failed records and resubmit only safe retry candidates; and set a clear evidence or confidence boundary and require review below it
- identify failed records and resubmit only safe retry candidates; and proceed whenever the assistant's prose sounds certain
Explanation
Correct Answer
C:In a tender-response data extractor, selective retry avoids paying again for already successful items; alongside this, risk-weighted gates limit autonomous actions when support is inadequate, satisfying the distinct requirements that a partially failed batch is repaired efficiently and that an uncertain high-impact decision is routed for review instead of automatic execution.
Incorrect Answers
A:In a tender-response data extractor, successful work would be duplicated unnecessarily and fluent language is not a calibrated reliability measure; the design would fail both conditions: a partially failed batch is repaired efficiently and an uncertain high-impact decision is routed for review instead of automatic execution.
B:In a tender-response data extractor, risk-weighted gates limit autonomous actions when support is inadequate, but successful work would be duplicated unnecessarily; therefore the separate requirement that a partially failed batch is repaired efficiently would remain unsatisfied.
D:In a tender-response data extractor, selective retry avoids paying again for already successful items, but fluent language is not a calibrated reliability measure; therefore the separate requirement that an uncertain high-impact decision is routed for review instead of automatic execution would remain unsatisfied.