Question 300 A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. Which design change directly satisfies the stated constraints? Treat the benchmark as insufficient, run an Indonesian task evaluation and prefer a model with proven target-language behavior. Select or route to a candidate validated on the actual target languages and legal terminology, even if its unit inference cost is higher. Choose a model supporting the required context and evaluate citation accuracy under measured full-context loads or redesign retrieval to fit. Benchmark smaller candidate models for classification accuracy and P95 latency, selecting the least costly model meeting both targets. ExplanationCorrect Answer B:The action targets the stated cause: Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. Check the repair by: compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. Other Answers A:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. C:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. D:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text.
Question 301 A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. Which causal explanation best accounts for the failure? A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. Model size does not determine whether a simple constrained task needs expensive general reasoning capacity. Model-card benchmark coverage does not automatically generalize to untested languages or task distributions. ExplanationCorrect Answer B:Select or route to a candidate validated on the actual target languages and legal terminology, even if its unit inference cost is higher. Other Answers A:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. C:That mechanism describes another problem: A team selects the largest available model for short classification messages, producing accurate labels but unacceptable latency and expense. It does not explain this case, where model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. D:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text.
Question 302 A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. Which concrete acceptance test best demonstrates the proposed change works? Compare response correctness on a labeled Indonesian set and document coverage gaps from the model card. Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. Run labeled classification and load tests with cost per thousand requests, macro-F1 and P95 latency. Count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text. ExplanationCorrect Answer B:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. Other Answers A:Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. C:Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. D:Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold.
Question 303 A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. Which operational practice should be made a durable requirement? Set maximum-context validation and retrieval fallback rules in the model-selection contract. Use task-specific quality, cost and latency gates instead of model-size preference. Maintain target-language evidence requirements and uncertainty flags during model selection. Maintain language-segmented evaluation and safe fallback for unsupported locales. ExplanationCorrect Answer D:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. Other Answers A:This control addresses a different exposure: A long-policy analysis requires 60k tokens of context, but the chosen low-cost model truncates the source and invents missing conclusions. It would not prevent this failure, because model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. B:That model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. C:Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text.
Question 304 A long-policy analysis requires 60k tokens of context, but the chosen low-cost model truncates the source and invents missing conclusions. What should the implementation team change first to resolve the described defect? Select or route to a candidate validated on the actual target languages and legal terminology, even if its unit inference cost is higher. Benchmark smaller candidate models for classification accuracy and P95 latency, selecting the least costly model meeting both targets. Choose a model supporting the required context and evaluate citation accuracy under measured full-context loads or redesign retrieval to fit. Select an embedding model whose input limit fits the chunk budget or split sections at meaningful boundaries before encoding. ExplanationCorrect Answer C:The action targets the stated cause: A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. Check the repair by: count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text. Other Answers A:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. B:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. D:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding.
Question 305 A long-policy analysis requires 60k tokens of context, but the chosen low-cost model truncates the source and invents missing conclusions. Why does the current approach fail under these conditions? Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. Model size does not determine whether a simple constrained task needs expensive general reasoning capacity. Embedding input truncation drops semantic evidence, creating vectors that omit later constraints and citations. A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. ExplanationCorrect Answer D:Choose a model supporting the required context and evaluate citation accuracy under measured full-context loads or redesign retrieval to fit. Other Answers A:That mechanism describes another problem: A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. It does not explain this case, where a model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. B:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. C:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding.
Question 306 A long-policy analysis requires 60k tokens of context, but the chosen low-cost model truncates the source and invents missing conclusions. Which observation would establish that the requirement is now met? Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. Compare source token counts with endpoint limits and query for details near the end of long sections. Count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text. Run labeled classification and load tests with cost per thousand requests, macro-F1 and P95 latency. ExplanationCorrect Answer C:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. Other Answers A:Count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text. B:Count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text. D:Count real prompt tokens on worst-case policies and verify final citations reference included—not truncated—text.
Question 307 A long-policy analysis requires 60k tokens of context, but the chosen low-cost model truncates the source and invents missing conclusions. What lasting control best prevents recurrence of the identified issue? Validate token lengths before embedding and maintain model-specific chunking profiles. Use task-specific quality, cost and latency gates instead of model-size preference. Set maximum-context validation and retrieval fallback rules in the model-selection contract. Maintain language-segmented evaluation and safe fallback for unsupported locales. ExplanationCorrect Answer C:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. Other Answers A:That a model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. B:A model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding. D:This control addresses a different exposure: A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. It would not prevent this failure, because a model's advertised context window constrains input and output budgeting; silently truncated evidence breaks grounding.
Question 308 An image-heavy manufacturing inspection app sends equipment photographs to a text-only endpoint, returning generic defect guesses. What is the most targeted corrective step for this observed failure? Benchmark smaller candidate models for classification accuracy and P95 latency, selecting the least costly model meeting both targets. Select an embedding model whose input limit fits the chunk budget or split sections at meaningful boundaries before encoding. Select or route to a candidate validated on the actual target languages and legal terminology, even if its unit inference cost is higher. Select a supported multimodal vision-capable model or an OCR/vision preprocessing stage before text reasoning, then evaluate defect recall. ExplanationCorrect Answer D:The action targets the stated cause: Text-only model inputs lack the image evidence required to identify visual fault features. Check the repair by: evaluate labeled inspection photographs and verify the output references actual observed components rather than invented findings. Other Answers A:Text-only model inputs lack the image evidence required to identify visual fault features. B:Text-only model inputs lack the image evidence required to identify visual fault features. C:Text-only model inputs lack the image evidence required to identify visual fault features.
Question 309 An image-heavy manufacturing inspection app sends equipment photographs to a text-only endpoint, returning generic defect guesses. What is the decisive mechanism behind the observed result? Model size does not determine whether a simple constrained task needs expensive general reasoning capacity. Embedding input truncation drops semantic evidence, creating vectors that omit later constraints and citations. Text-only model inputs lack the image evidence required to identify visual fault features. Model language coverage and domain performance vary; English-only scores do not predict accuracy on Japanese legal text. ExplanationCorrect Answer C:Select a supported multimodal vision-capable model or an OCR/vision preprocessing stage before text reasoning, then evaluate defect recall. Other Answers A:Text-only model inputs lack the image evidence required to identify visual fault features. B:Text-only model inputs lack the image evidence required to identify visual fault features. D:That mechanism describes another problem: A multilingual legal assistant's inexpensive model handles English but fails on Japanese clauses that affect customer obligations. It does not explain this case, where text-only model inputs lack the image evidence required to identify visual fault features.
Question 310 An image-heavy manufacturing inspection app sends equipment photographs to a text-only endpoint, returning generic defect guesses. Which verification is strongest for this particular failure mode? Evaluate labeled inspection photographs and verify the output references actual observed components rather than invented findings. Compare clause-extraction errors by language and ensure the selected model meets each locale's safety threshold. Compare source token counts with endpoint limits and query for details near the end of long sections. Run labeled classification and load tests with cost per thousand requests, macro-F1 and P95 latency. ExplanationCorrect Answer A:Text-only model inputs lack the image evidence required to identify visual fault features. Other Answers B:Evaluate labeled inspection photographs and verify the output references actual observed components rather than invented findings. C:Evaluate labeled inspection photographs and verify the output references actual observed components rather than invented findings. D:Evaluate labeled inspection photographs and verify the output references actual observed components rather than invented findings.
