Question 680 An assistant cites a document whose passage does not contain the asserted procedure. What should the engineer choose? Use citation correctness tests on source-linked answers Test prompt and context size limits on worst-case queries Measure retrieval recall separately from generation quality Evaluate refusal and escalation on sensitive unsupported requests Explanation Correct Answer A:This approach ensures a displayed reference actually supports the claim. Incorrect Answers B:This choice prevents truncation and silent context loss. This does not meet the need for a method that ensures a displayed reference actually supports the claim. C:That technique identifies whether failures originate in search or model reasoning. This does not meet the need for a method that ensures a displayed reference actually supports the claim. D:This alternative reduces high-impact unsupported guidance. This does not meet the need for a method that ensures a displayed reference actually supports the claim.
Question 681 A RAG deployment must rotate credentials without rewriting business queries. What should the engineer choose? Move slow model calls outside long-running SQL transactions Use managed identity or approved credential isolation for external model calls Invoke an approved external REST endpoint from supported SQL using sp_invoke_external_rest_endpoint Validate and parse returned model JSON before downstream database writes Explanation Correct Answer B:This implementation avoids plaintext secrets in stored procedures and deployment scripts. Incorrect Answers A:This alternative avoids holding row locks while waiting for network inference. This does not meet the need for a method that avoids plaintext secrets in stored procedures and deployment scripts. C:This choice integrates supported SQL workloads with external model APIs. This does not meet the need for a method that avoids plaintext secrets in stored procedures and deployment scripts. D:This alternative prevents malformed or invented model fields from corrupting data. This does not meet the need for a method that avoids plaintext secrets in stored procedures and deployment scripts.
Question 682 A chatbot's retrieved snippets contradict each other on a key obligation. Which approach would you recommend? Include source identifiers and relevant excerpts with generated answers Require a no-answer fallback when retrieved evidence is insufficient Use retrieval-augmented generation for answers grounded in changing governed data Limit retrieved context to relevant chunks within token budget Explanation Correct Answer B:Require a no-answer fallback when retrieved evidence is insufficient fits because it avoids unsupported fabricated claims. That addresses the stated technical constraint. Incorrect Answers A:That technique makes claims traceable to retrieved passages. This does not meet the need for a method that avoids unsupported fabricated claims. C:That technique injects authorized current context into prompts before generation. This does not meet the need for a method that avoids unsupported fabricated claims. D:This option prevents excessive noise and unnecessary model cost. This does not meet the need for a method that avoids unsupported fabricated claims.
Question 683 A medical portal must not retrieve another patient's confidential notes. The SQL deployment supports the named external model and REST integration. What is the most appropriate design decision? Guard retrieved documents against prompt-injection instructions Retrieve only the sources the caller is authorized to access Limit retrieved context to relevant chunks within token budget Use retrieval-augmented generation for answers grounded in changing governed data Explanation Correct Answer B:This approach prevents RAG from bypassing document-level permissions. Incorrect Answers A:This choice treats database content as data rather than high-priority commands. This does not meet the need for a method that prevents RAG from bypassing document-level permissions. C:This option prevents excessive noise and unnecessary model cost. This does not meet the need for a method that prevents RAG from bypassing document-level permissions. D:This option injects authorized current context into prompts before generation. This does not meet the need for a method that prevents RAG from bypassing document-level permissions.
Question 684 A stored procedure receives model output that may omit required properties. Which change most directly meets the requirement? Set bounded timeout and retry behavior for SQL-external calls Invoke an approved external REST endpoint from supported SQL using sp_invoke_external_rest_endpoint Move slow model calls outside long-running SQL transactions Validate and parse returned model JSON before downstream database writes Explanation Correct Answer D:Validate and parse returned model JSON before downstream database writes fits because it prevents malformed or invented model fields from corrupting data. That addresses the stated technical constraint. Incorrect Answers A:That technique protects database resources when models are slow or unavailable. This does not meet the need for a method that prevents malformed or invented model fields from corrupting data. B:This alternative integrates supported SQL workloads with external model APIs. This does not meet the need for a method that prevents malformed or invented model fields from corrupting data. C:That technique avoids holding row locks while waiting for network inference. This does not meet the need for a method that prevents malformed or invented model fields from corrupting data.
Question 685 A RAG workflow sends a few vetted product specifications to an external model. Which solution best satisfies the stated constraint? Use managed identity or approved credential isolation for external model calls Convert selected SQL rows to a compact JSON payload for model context Invoke an approved external REST endpoint from supported SQL using sp_invoke_external_rest_endpoint Validate and parse returned model JSON before downstream database writes Explanation Correct Answer B:This implementation structures authorized database facts in a model-consumable form. Incorrect Answers A:This option avoids plaintext secrets in stored procedures and deployment scripts. This does not meet the need for a method that structures authorized database facts in a model-consumable form. C:This choice integrates supported SQL workloads with external model APIs. This does not meet the need for a method that structures authorized database facts in a model-consumable form. D:This alternative prevents malformed or invented model fields from corrupting data. This does not meet the need for a method that structures authorized database facts in a model-consumable form.
Question 686 A service hits context-length limits because it includes all search hits. Which solution best satisfies the stated constraint? Use retrieval-augmented generation for answers grounded in changing governed data Include source identifiers and relevant excerpts with generated answers Limit retrieved context to relevant chunks within token budget Guard retrieved documents against prompt-injection instructions Explanation Correct Answer C:This approach prevents excessive noise and unnecessary model cost. Incorrect Answers A:This option injects authorized current context into prompts before generation. This does not meet the need for a method that prevents excessive noise and unnecessary model cost. B:This choice makes claims traceable to retrieved passages. This does not meet the need for a method that prevents excessive noise and unnecessary model cost. D:This alternative treats database content as data rather than high-priority commands. This does not meet the need for a method that prevents excessive noise and unnecessary model cost.
Question 687 A RAG service truncates the most recent policy because too many chunks were included. The SQL deployment supports the named external model and REST integration. Which solution is most directly applicable? Test prompt and context size limits on worst-case queries Use citation correctness tests on source-linked answers Evaluate answer faithfulness against retrieved source passages Measure retrieval recall separately from generation quality Explanation Correct Answer A:Test prompt and context size limits on worst-case queries fits because it prevents truncation and silent context loss. That addresses the stated technical constraint. Incorrect Answers B:This option ensures a displayed reference actually supports the claim. This does not meet the need for a method that prevents truncation and silent context loss. C:This option detects claims unsupported by context even when fluent. This does not meet the need for a method that prevents truncation and silent context loss. D:That technique identifies whether failures originate in search or model reasoning. This does not meet the need for a method that prevents truncation and silent context loss.
Question 688 A retrieved customer comment attempts to exfiltrate secrets via the model response. What should the SQL developer implement? Limit retrieved context to relevant chunks within token budget Use retrieval-augmented generation for answers grounded in changing governed data Include source identifiers and relevant excerpts with generated answers Guard retrieved documents against prompt-injection instructions Explanation Correct Answer D:Guard retrieved documents against prompt-injection instructions fits because it treats database content as data rather than high-priority commands. That addresses the stated technical constraint. Incorrect Answers A:This choice prevents excessive noise and unnecessary model cost. This does not meet the need for a method that treats database content as data rather than high-priority commands. B:This choice injects authorized current context into prompts before generation. This does not meet the need for a method that treats database content as data rather than high-priority commands. C:That technique makes claims traceable to retrieved passages. This does not meet the need for a method that treats database content as data rather than high-priority commands.
Question 689 An external AI request blocks a transaction holding customer records. Which approach would you recommend? Move slow model calls outside long-running SQL transactions Validate and parse returned model JSON before downstream database writes Invoke an approved external REST endpoint from supported SQL using sp_invoke_external_rest_endpoint Convert selected SQL rows to a compact JSON payload for model context Explanation Correct Answer A:Move slow model calls outside long-running SQL transactions is the appropriate choice: it avoids holding row locks while waiting for network inference. The deployment should be tested with representative data. Incorrect Answers B:This choice prevents malformed or invented model fields from corrupting data. This does not meet the need for a method that avoids holding row locks while waiting for network inference. C:This choice integrates supported SQL workloads with external model APIs. This does not meet the need for a method that avoids holding row locks while waiting for network inference. D:That technique structures authorized database facts in a model-consumable form. This does not meet the need for a method that avoids holding row locks while waiting for network inference.
Question 690 An index update changes retrieved passages and incidents require traceability. What is the most appropriate design decision? Track model and retrieval configuration versions for production answers Evaluate answer faithfulness against retrieved source passages Test prompt and context size limits on worst-case queries Measure retrieval recall separately from generation quality Explanation Correct Answer A:This implementation enables reproducible investigations of changing assistant behavior. Incorrect Answers B:That technique detects claims unsupported by context even when fluent. This does not meet the need for a method that enables reproducible investigations of changing assistant behavior. C:That technique prevents truncation and silent context loss. This does not meet the need for a method that enables reproducible investigations of changing assistant behavior. D:This alternative identifies whether failures originate in search or model reasoning. This does not meet the need for a method that enables reproducible investigations of changing assistant behavior.