Question 230 A customer-service assistant gives a helpful policy summary but uses harassing language toward the requester in its final paragraph. Which design change directly satisfies the stated constraints? Report sampled monitoring coverage and uncertainty separately from full-bank offline evaluation or qualified human review. Treat the response as incomplete and add procedure-coverage checks that require the safety prerequisite before any repair instructions. Mark the result as unsupported, require claim-to-passage grounding and abstain or escalate when the source lacks the needed limit. Fail the safety and tone criterion independently of policy factuality and route the case through controlled moderation or rewrite review. ExplanationCorrect Answer D:The action targets the stated cause: Correct factual content does not make abusive presentation acceptable; quality evaluation needs separate safety dimensions. Check the repair by: score factuality and harmful-language behavior separately on adversarial tone prompts and confirm safe alternatives pass. Other Answers A:Correct factual content does not make abusive presentation acceptable; quality evaluation needs separate safety dimensions. B:Correct factual content does not make abusive presentation acceptable; quality evaluation needs separate safety dimensions. C:Correct factual content does not make abusive presentation acceptable; quality evaluation needs separate safety dimensions.
Question 231 Long, broad chunks yield a high retrieval hit rate but the agent cites unrelated sections for narrow warranty questions. Which implementation decision best restores the intended behavior? Recalibrate retrieval thresholds on labeled positives and negatives before promoting the new embedding model. Compare smaller section-aware chunks and a lower top-k using grounded-answer evaluations, selecting the configuration that improves citation precision. Add selective reranking of a modest candidate pool and keep a fixed context budget rather than sending all fifteen passages. Increase chunk boundary awareness or bounded overlap so prerequisites and actions remain together for calibration queries. ExplanationCorrect Answer B:The action targets the stated cause: Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. Check the repair by: evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations. Other Answers A:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. C:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. D:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context.
Question 232 Long, broad chunks yield a high retrieval hit rate but the agent cites unrelated sections for narrow warranty questions. Which underlying technical principle explains why this occurs? Similarity scales and distributions depend on the embedding model and cannot be assumed stable across versions. Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. ExplanationCorrect Answer D:Compare smaller section-aware chunks and a lower top-k using grounded-answer evaluations, selecting the configuration that improves citation precision. Other Answers A:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. B:That mechanism describes another problem: A hybrid search retrieves a relevant policy at rank 15 while the generator sees only the top five chunks; increasing context size harms latency. It does not explain this case, where coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. C:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context.
Question 233 Long, broad chunks yield a high retrieval hit rate but the agent cites unrelated sections for narrow warranty questions. After correction, what targeted test should demonstrate success? Compare answer support and P95 latency with and without reranking on the same labeled query set. Evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations. Plot positive/negative score distributions, adjust the threshold and verify recall plus false-positive rates. Inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set. ExplanationCorrect Answer B:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. Other Answers A:Evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations. C:Evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations. D:Evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations.
Question 234 Long, broad chunks yield a high retrieval hit rate but the agent cites unrelated sections for narrow warranty questions. Which release safeguard most directly addresses this failure class? Treat embedding-model upgrades as re-indexing and evaluation events with threshold regression tests. Track source-level evidence quality, not only raw hit counts, when tuning retrieval. Include cross-sentence dependency questions in the retrieval regression dataset. Establish separate ranking and latency acceptance gates for retrieval changes. ExplanationCorrect Answer B:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. Other Answers A:This control addresses a different exposure: A new embedding model changes similarity scores, and a fixed threshold rejects most valid questions even though the source corpus is unchanged. It would not prevent this failure, because coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. C:That coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. D:Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context.
Question 235 Tiny sentence chunks improve keyword precision but make an instruction about calibration useless because its required prerequisite sits one sentence earlier. Which design change directly satisfies the stated constraints? Increase chunk boundary awareness or bounded overlap so prerequisites and actions remain together for calibration queries. Compare smaller section-aware chunks and a lower top-k using grounded-answer evaluations, selecting the configuration that improves citation precision. Add selective reranking of a modest candidate pool and keep a fixed context budget rather than sending all fifteen passages. Rebalance metadata filters and lexical/vector hybrid search for exact code matches while measuring answer support. ExplanationCorrect Answer A:The action targets the stated cause: Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. Check the repair by: inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set. Other Answers B:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. C:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. D:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations.
Question 236 Tiny sentence chunks improve keyword precision but make an instruction about calibration useless because its required prerequisite sits one sentence earlier. Which causal explanation best accounts for the failure? Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. Pure semantic ranking can underweight rare exact identifiers when corpus density changes. Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. ExplanationCorrect Answer B:Increase chunk boundary awareness or bounded overlap so prerequisites and actions remain together for calibration queries. Other Answers A:That mechanism describes another problem: Long, broad chunks yield a high retrieval hit rate but the agent cites unrelated sections for narrow warranty questions. It does not explain this case, where excessively fine segmentation removes procedural dependencies necessary for grounded explanations. C:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. D:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations.
Question 237 Tiny sentence chunks improve keyword precision but make an instruction about calibration useless because its required prerequisite sits one sentence earlier. Which concrete acceptance test best demonstrates the proposed change works? Compare answer support and P95 latency with and without reranking on the same labeled query set. Inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set. Evaluate held-out warranty questions for support span accuracy and retrieval relevance across both chunking configurations. Query held-out part numbers and compare exact-ID recall and groundedness before and after hybrid filtering. ExplanationCorrect Answer B:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. Other Answers A:Inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set. C:Inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set. D:Inspect retrieved calibration answers and verify prerequisites and measurement steps are present in a single sufficient evidence set.
Question 238 Tiny sentence chunks improve keyword precision but make an instruction about calibration useless because its required prerequisite sits one sentence earlier. Which operational practice should be made a durable requirement? Track source-level evidence quality, not only raw hit counts, when tuning retrieval. Establish separate ranking and latency acceptance gates for retrieval changes. Include cross-sentence dependency questions in the retrieval regression dataset. Monitor identifier coverage and apply corpus-drift checks when ingestion volume changes. ExplanationCorrect Answer C:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. Other Answers A:That excessively fine segmentation removes procedural dependencies necessary for grounded explanations. B:Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. D:This control addresses a different exposure: A document collection adds thousands of unrelated records, and previously good retrieval now shows poor grounding for rare product codes. It would not prevent this failure, because excessively fine segmentation removes procedural dependencies necessary for grounded explanations.
Question 239 A hybrid search retrieves a relevant policy at rank 15 while the generator sees only the top five chunks; increasing context size harms latency. What should the implementation team change first to resolve the described defect? Increase chunk boundary awareness or bounded overlap so prerequisites and actions remain together for calibration queries. Add selective reranking of a modest candidate pool and keep a fixed context budget rather than sending all fifteen passages. Compare smaller section-aware chunks and a lower top-k using grounded-answer evaluations, selecting the configuration that improves citation precision. Choose a model supporting the required context and evaluate citation accuracy under measured full-context loads or redesign retrieval to fit. ExplanationCorrect Answer B:The action targets the stated cause: Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. Check the repair by: compare answer support and P95 latency with and without reranking on the same labeled query set. Other Answers A:Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. C:Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. D:Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion.
Question 240 A hybrid search retrieves a relevant policy at rank 15 while the generator sees only the top five chunks; increasing context size harms latency. Why does the current approach fail under these conditions? Coarse chunks can match topic vocabulary yet bury the actual answer span among unrelated context. A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. Excessively fine segmentation removes procedural dependencies necessary for grounded explanations. ExplanationCorrect Answer C:Add selective reranking of a modest candidate pool and keep a fixed context budget rather than sending all fifteen passages. Other Answers A:Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. B:Candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion. D:That mechanism describes another problem: Tiny sentence chunks improve keyword precision but make an instruction about calibration useless because its required prerequisite sits one sentence earlier. It does not explain this case, where candidate recall may be adequate while final ranking fails; reranking addresses ordering without unbounded prompt expansion.
