100% Latest & Updated Microsoft DP-800 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 940
An AI assistant answers questions from frequently changing internal procedures. A test instead finds: Assistant answers with last year's retirement policy. Which underlying issue best explains this finding?
- RAG context construction over-shared database columns
- Context included duplicate conflicting revisions
- Retrieval included stale records or skipped freshness filter
- LLM inferred inventory from stale examples
Explanation
Correct Answer
C:Retrieval included stale records or skipped freshness filter. It accounts for the specific observed failure: assistant answers with last year's retirement policy. The intended solution is use rag to ground generation in current approved records with the required setup.
Incorrect Answers
A:RAG context construction over-shared database columns. This concerns external model receives unrelated customer details. Here, retrieval included stale records or skipped freshness filter.
B:Context included duplicate conflicting revisions. This concerns model quotes obsolete value despite new approved record. Here, retrieval included stale records or skipped freshness filter.
D:LLM inferred inventory from stale examples. This concerns assistant offers unavailable quantity despite empty stock. Here, retrieval included stale records or skipped freshness filter.
Question 941
An LLM must not invent answer when SQL retrieval returns no supporting rows. Which database design or engineering approach most directly meets this requirement?
- Limit top-K and trim passages by relevance
- Define prompt template with instructions and bounded evidence section
- Use evidence-aware abstention in RAG prompting
- Use approved Foundry endpoint and managed authentication
Explanation
Correct Answer
C:Use evidence-aware abstention in RAG prompting. The implementation is explicitly instruct refusal or escalation when context lacks answer; success can be checked by test out-of-scope question returns grounded uncertainty. The method addresses the stated data and workload constraint.
Incorrect Answers
A:Limit top-K and trim passages by relevance. This concerns a SQL-backed assistant runs retrieval but sends too many irrelevant rows. Instead, use evidence-aware abstention in RAG prompting.
B:Define prompt template with instructions and bounded evidence section. This concerns a reusable prompt with SQL database retrieval. Instead, use evidence-aware abstention in RAG prompting.
D:Use approved Foundry endpoint and managed authentication. This concerns connecting SQL retrieval to microsoft foundry model. Instead, use evidence-aware abstention in RAG prompting.
Question 942
An LLM must not invent answer when SQL retrieval returns no supporting rows. The team selects: Use evidence-aware abstention in RAG prompting. Which implementation action puts this design into practice?
- Explicitly instruct refusal or escalation when context lacks answer
- Build structured payload of question, context and citation IDs
- Configure endpoint address, credentials and model deployment
- Generate hybrid top-K with RRF and bounded source context
Explanation
Correct Answer
A:Explicitly instruct refusal or escalation when context lacks answer. This is the concrete setup for use evidence-aware abstention in rag prompting, and test out-of-scope question returns grounded uncertainty supplies a direct confirmation.
Incorrect Answers
B:Build structured payload of question, context and citation IDs. This concerns a reusable prompt with SQL database retrieval. Here, explicitly instruct refusal or escalation when context lacks answer.
C:Configure endpoint address, credentials and model deployment. This concerns connecting SQL retrieval to microsoft foundry model. Here, explicitly instruct refusal or escalation when context lacks answer.
D:Generate hybrid top-K with RRF and bounded source context. This concerns hybrid SQL/vector candidates before generating answer. Here, explicitly instruct refusal or escalation when context lacks answer.
Question 943
An LLM must not invent answer when SQL retrieval returns no supporting rows. After implementing the design, which check most directly validates the expected behavior?
- Measure answer quality and input token count
- Test out-of-scope question returns grounded uncertainty
- Run test prompt and verify output plus usage metrics
- Test authorized response and telemetry
Explanation
Correct Answer
B:Test out-of-scope question returns grounded uncertainty. This verifies the requirement in the case rather than merely confirming that a job ran; an opposing result can reveal that prompt lacks no-evidence behavior or model ignored it.
Incorrect Answers
A:Measure answer quality and input token count. This concerns a SQL-backed assistant runs retrieval but sends too many irrelevant rows. Instead, test out-of-scope question returns grounded uncertainty.
C:Run test prompt and verify output plus usage metrics. This concerns connecting SQL retrieval to microsoft foundry model. Instead, test out-of-scope question returns grounded uncertainty.
D:Test authorized response and telemetry. This concerns a stored procedure sends an allowed request to an external text model. Instead, test out-of-scope question returns grounded uncertainty.
Question 944
An LLM must not invent answer when SQL retrieval returns no supporting rows. A test instead finds: Assistant confidently makes up non-existent product policy. Which underlying issue best explains this finding?
- Endpoint authentication or credential configuration is missing
- Prompt lacks no-evidence behavior or model ignored it
- SQL rows were flattened without structured context
- Endpoint or model deployment name was incorrect
Explanation
Correct Answer
B:Prompt lacks no-evidence behavior or model ignored it. It accounts for the specific observed failure: assistant confidently makes up non-existent product policy. The intended solution is use evidence-aware abstention in rag prompting with the required setup.
Incorrect Answers
A:Endpoint authentication or credential configuration is missing. This concerns external rest call fails with unauthorized response. Here, prompt lacks no-evidence behavior or model ignored it.
C:SQL rows were flattened without structured context. This concerns llm receives ambiguous comma-separated values. Here, prompt lacks no-evidence behavior or model ignored it.
D:Endpoint or model deployment name was incorrect. This concerns rest request hits wrong model deployment. Here, prompt lacks no-evidence behavior or model ignored it.
Question 945
A stored procedure sends an allowed request to an external text model. Which database design or engineering approach most directly meets this requirement?
- Send only minimum necessary approved fields
- Serialize controlled result fields to JSON
- Use approved Foundry endpoint and managed authentication
- Use sp_invoke_external_rest_endpoint where supported and configured
Explanation
Correct Answer
D:Use sp_invoke_external_rest_endpoint where supported and configured. The implementation is create scoped credential/endpoint and call stored procedure with json payload; success can be checked by test authorized response and telemetry. The method addresses the stated data and workload constraint.
Incorrect Answers
A:Send only minimum necessary approved fields. This concerns a support bot routes answers to an external model via rest. Instead, use sp_invoke_external_rest_endpoint where supported and configured.
B:Serialize controlled result fields to JSON. This is relevant when a chatbot uses SQL result rows as context for a model. Instead, use sp_invoke_external_rest_endpoint where supported and configured.
C:Use approved Foundry endpoint and managed authentication. This concerns connecting SQL retrieval to microsoft foundry model. Instead, use sp_invoke_external_rest_endpoint where supported and configured.
Question 946
A stored procedure sends an allowed request to an external text model. The team selects: Use sp_invoke_external_rest_endpoint where supported and configured. Which implementation action puts this design into practice?
- Apply RLS predicate and per-document ACL before prompt construction
- Use JSON_VALUE or OPENJSON with schema checks
- Create SQL projection excluding secrets before external request
- Create scoped credential/endpoint and call stored procedure with JSON payload
Explanation
Correct Answer
D:Create scoped credential/endpoint and call stored procedure with JSON payload. This is the concrete setup for use sp_invoke_external_rest_endpoint where supported and configured, and test authorized response and telemetry supplies a direct confirmation.
Incorrect Answers
A:Apply RLS predicate and per-document ACL before prompt construction. This concerns rag that respects row-level security. Instead, create scoped credential/endpoint and call stored procedure with JSON payload.
B:Use JSON_VALUE or OPENJSON with schema checks. This is relevant when an assistant produces a json answer that must be stored in SQL. Instead, create scoped credential/endpoint and call stored procedure with JSON payload.
C:Create SQL projection excluding secrets before external request. This concerns a support bot routes answers to an external model via rest. Instead, create scoped credential/endpoint and call stored procedure with JSON payload.
Question 947
A stored procedure sends an allowed request to an external text model. After implementing the design, which check most directly validates the expected behavior?
- Run test prompt and verify output plus usage metrics
- Test authorized response and telemetry
- Inspect outbound JSON for PII and hidden columns
- Reject malformed response and accept valid typed output
Explanation
Correct Answer
B:Test authorized response and telemetry. This verifies the requirement in the case rather than merely confirming that a job ran; an opposing result can reveal that endpoint authentication or credential configuration is missing.
Incorrect Answers
A:Run test prompt and verify output plus usage metrics. This concerns connecting SQL retrieval to microsoft foundry model. Instead, test authorized response and telemetry.
C:Inspect outbound JSON for PII and hidden columns. This concerns a support bot routes answers to an external model via rest. Instead, test authorized response and telemetry.
D:Reject malformed response and accept valid typed output. This is relevant when an assistant produces a json answer that must be stored in SQL. Instead, test authorized response and telemetry.
Question 948
A stored procedure sends an allowed request to an external text model. A test instead finds: External REST call fails with unauthorized response. Which underlying issue best explains this finding?
- Endpoint authentication or credential configuration is missing
- RAG context construction over-shared database columns
- Endpoint or model deployment name was incorrect
- Context was flooded with low-relevance passages
Explanation
Correct Answer
A:Endpoint authentication or credential configuration is missing. It accounts for the specific observed failure: external rest call fails with unauthorized response. The intended solution is use sp_invoke_external_rest_endpoint where supported and configured with the required setup.
Incorrect Answers
B:RAG context construction over-shared database columns. This concerns external model receives unrelated customer details. Here, endpoint authentication or credential configuration is missing.
C:Endpoint or model deployment name was incorrect. This concerns rest request hits wrong model deployment. Here, endpoint authentication or credential configuration is missing.
D:Context was flooded with low-relevance passages. This concerns token cost spikes without improved grounding. Here, endpoint authentication or credential configuration is missing.
Question 949
A chatbot uses SQL result rows as context for a model. Which database design or engineering approach most directly meets this requirement?
- Serialize controlled result fields to JSON
- Limit top-K and trim passages by relevance
- Send only minimum necessary approved fields
- Validate and extract model JSON response before persistence
Explanation
Correct Answer
A:Serialize controlled result fields to JSON. The implementation is use for json path or json_object projection of allowed fields; success can be checked by parse payload and verify required field names and types. The method addresses the stated data and workload constraint.
Incorrect Answers
B:Limit top-K and trim passages by relevance. This concerns a SQL-backed assistant runs retrieval but sends too many irrelevant rows. Here, serialize controlled result fields to JSON.
C:Send only minimum necessary approved fields. This concerns a support bot routes answers to an external model via rest. Here, serialize controlled result fields to JSON.
D:Validate and extract model JSON response before persistence. This is relevant when an assistant produces a json answer that must be stored in SQL. Here, serialize controlled result fields to JSON.
Question 950
A chatbot uses SQL result rows as context for a model. The team selects: Serialize controlled result fields to JSON. Which implementation action puts this design into practice?
- Use JSON_VALUE or OPENJSON with schema checks
- Create scoped credential/endpoint and call stored procedure with JSON payload
- Use FOR JSON PATH or JSON_OBJECT projection of allowed fields
- Configure endpoint address, credentials and model deployment
Explanation
Correct Answer
C:Use FOR JSON PATH or JSON_OBJECT projection of allowed fields. This is the concrete setup for serialize controlled result fields to json, and parse payload and verify required field names and types supplies a direct confirmation.
Incorrect Answers
A:Use JSON_VALUE or OPENJSON with schema checks. This is relevant when an assistant produces a json answer that must be stored in SQL. Instead, use FOR JSON PATH or JSON_OBJECT projection of allowed fields.
B:Create scoped credential/endpoint and call stored procedure with JSON payload. This concerns a stored procedure sends an allowed request to an external text model. Instead, use FOR JSON PATH or JSON_OBJECT projection of allowed fields.
D:Configure endpoint address, credentials and model deployment. This concerns connecting SQL retrieval to microsoft foundry model. Instead, use FOR JSON PATH or JSON_OBJECT projection of allowed fields.
Developing AI-Enabled Database Solutions (DP-800) Practice Tests