Question 270 An assistant repeatedly hallucinates product specifications when asked “tell me everything,” despite having retrieved product sheets. Which verification is strongest for this particular failure mode? Diff translated outputs against source commands and part numbers on a multilingual regression corpus. Replay two currencies and verify quoted amounts reconcile with the authorized rate and product source. Compare outputs on a held-out incident set for length, correct facts and a clearly identified next action. Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements. ExplanationCorrect Answer D:General prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. Other Answers A:Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements. B:Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements. C:Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements.
Question 271 An assistant repeatedly hallucinates product specifications when asked “tell me everything,” despite having retrieved product sheets. What change to engineering governance would reduce this risk over time? Measure groundedness on missing-field and conflicting-source examples during every prompt revision. Maintain protected-token extraction and post-translation equality checks for technical content. Keep target-style exemplars and rubric-based prompt regression tests for the incident role. Use typed context enrichment and deterministic money calculations rather than prompt-only currency transformations. ExplanationCorrect Answer A:General prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. Other Answers B:That general prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. C:General prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. D:This control addresses a different exposure: A multilingual sales assistant extracts a product name but ignores the user's requested currency when preparing a quotation. It would not prevent this failure, because general prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive.
Question 272 A prompt asks the model to translate technical instructions, but it changes command-line flags and model numbers. Which engineering action most directly addresses this specific requirement? Require use of supplied specification fields, forbid unsupported values and include an explicit not_found response for missing attributes. Rewrite the instruction with a precise audience, length, output fields and evidence requirements, then evaluate concise outputs on realistic incidents. Add explicit immutable-token preservation rules with input/output checks for commands, identifiers and measurements. 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: Language translation should transform prose while preserving tokens that are meaningful machine identifiers. Check the repair by: diff translated outputs against source commands and part numbers on a multilingual regression corpus. Other Answers A:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. B:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. D:Language translation should transform prose while preserving tokens that are meaningful machine identifiers.
Question 273 A prompt asks the model to translate technical instructions, but it changes command-line flags and model numbers. What architectural constraint is most directly responsible? Open-ended prompts invite variable format and coverage; a concrete contract aligns generation with the operational consumer. General prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. Language translation should transform prose while preserving tokens that are meaningful machine identifiers. Embedding input truncation drops semantic evidence, creating vectors that omit later constraints and citations. ExplanationCorrect Answer C:Add explicit immutable-token preservation rules with input/output checks for commands, identifiers and measurements. Other Answers A:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. B:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. D:That mechanism describes another problem: Technical manuals contain 5,000-token sections, but the selected embedding endpoint accepts only much shorter inputs and silently truncates endings. It does not explain this case, where language translation should transform prose while preserving tokens that are meaningful machine identifiers.
Question 274 A prompt asks the model to translate technical instructions, but it changes command-line flags and model numbers. What validation evidence would most specifically confirm the resolution? Diff translated outputs against source commands and part numbers on a multilingual regression corpus. Compare source token counts with endpoint limits and query for details near the end of long sections. Compare outputs on a held-out incident set for length, correct facts and a clearly identified next action. Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements. ExplanationCorrect Answer A:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. Other Answers B:Diff translated outputs against source commands and part numbers on a multilingual regression corpus. C:Diff translated outputs against source commands and part numbers on a multilingual regression corpus. D:Diff translated outputs against source commands and part numbers on a multilingual regression corpus.
Question 275 A prompt asks the model to translate technical instructions, but it changes command-line flags and model numbers. What prevention measure is most targeted to this underlying cause? Maintain protected-token extraction and post-translation equality checks for technical content. Measure groundedness on missing-field and conflicting-source examples during every prompt revision. Validate token lengths before embedding and maintain model-specific chunking profiles. Keep target-style exemplars and rubric-based prompt regression tests for the incident role. ExplanationCorrect Answer A:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. Other Answers B:Language translation should transform prose while preserving tokens that are meaningful machine identifiers. C:This control addresses a different exposure: Technical manuals contain 5,000-token sections, but the selected embedding endpoint accepts only much shorter inputs and silently truncates endings. It would not prevent this failure, because language translation should transform prose while preserving tokens that are meaningful machine identifiers. D:That language translation should transform prose while preserving tokens that are meaningful machine identifiers.
Question 276 A model chooses an answer from four policy options but returns paragraphs of explanation where a downstream parser requires an option letter. Which implementation decision best restores the intended behavior? Specify a supported structured response containing answer_letter and short_rationale with an allowed-value enum, then validate parsing. Rewrite the instruction with a precise audience, length, output fields and evidence requirements, then evaluate concise outputs on realistic incidents. Gate side-effect tools with explicit allowed actions, resource-level authorization and confirmation before the call executes. Require use of supplied specification fields, forbid unsupported values and include an explicit not_found response for missing attributes. ExplanationCorrect Answer A:The action targets the stated cause: Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. Check the repair by: feed varied questions and verify the output always parses to exactly one allowed letter plus rationale. Other Answers B:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. C:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. D:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape.
Question 277 A model chooses an answer from four policy options but returns paragraphs of explanation where a downstream parser requires an option letter. Which underlying technical principle explains why this occurs? Final-response safety filters run too late to prevent a destructive tool invocation that has already occurred. Open-ended prompts invite variable format and coverage; a concrete contract aligns generation with the operational consumer. General prompts allow parametric guesses to compete with retrieved evidence; instruction hierarchy must make source-grounded facts decisive. Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. ExplanationCorrect Answer D:Specify a supported structured response containing answer_letter and short_rationale with an allowed-value enum, then validate parsing. Other Answers A:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. B:That mechanism describes another problem: A model gives a long narrative when an incident commander needs a two-sentence status update and one explicit next action. It does not explain this case, where examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. C:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape.
Question 278 A model chooses an answer from four policy options but returns paragraphs of explanation where a downstream parser requires an option letter. After correction, what targeted test should demonstrate success? Replay a destructive prompt and inspect that the attempted tool call is blocked before execution and recorded for review. Feed varied questions and verify the output always parses to exactly one allowed letter plus rationale. Compare outputs on a held-out incident set for length, correct facts and a clearly identified next action. Test deliberately absent specifications and verify the answer says unavailable instead of supplying invented measurements. ExplanationCorrect Answer B:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. Other Answers A:Feed varied questions and verify the output always parses to exactly one allowed letter plus rationale. C:Feed varied questions and verify the output always parses to exactly one allowed letter plus rationale. D:Feed varied questions and verify the output always parses to exactly one allowed letter plus rationale.
Question 279 A model chooses an answer from four policy options but returns paragraphs of explanation where a downstream parser requires an option letter. Which release safeguard most directly addresses this failure class? Separate read and write permissions and require server-side confirmation policies for destructive operations. Version prompt and output schema together, including malformed-output test cases. Measure groundedness on missing-field and conflicting-source examples during every prompt revision. Keep target-style exemplars and rubric-based prompt regression tests for the incident role. ExplanationCorrect Answer B:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. Other Answers A:This control addresses a different exposure: An assistant's final answer is safe, but its tool-call arguments contain an unauthorized delete operation generated from a user's “clean up” prompt. It would not prevent this failure, because examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. C:That examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape. D:Examples alone do not guarantee a machine-readable field contract; an enum and parser constrain the response shape.
Question 280 A retrieved contract includes a paragraph instructing an agent to send the customer's credentials to a web endpoint. Which design change directly satisfies the stated constraints? Add explicit immutable-token preservation rules with input/output checks for commands, identifiers and measurements. Redact or suppress sensitive values before tracing and use approved metadata-only observability for protected fields. Apply scope guardrails requiring educational information, uncertainty and referral for clinical decisions, and avoid actionable patient-specific prescriptions. Ignore document-embedded commands, constrain tool calls to server-side permissions and redact credentials from the model context. ExplanationCorrect Answer D:The action targets the stated cause: Retrieved passages are untrusted task data, not higher-priority instructions authorizing external actions. Check the repair by: inject the malicious paragraph in a test corpus and confirm no network tool receives credential values. Other Answers A:Retrieved passages are untrusted task data, not higher-priority instructions authorizing external actions. B:Retrieved passages are untrusted task data, not higher-priority instructions authorizing external actions. C:Retrieved passages are untrusted task data, not higher-priority instructions authorizing external actions.
