Question 470
The engineering change request is to generate embeddings during indexing with a skillset. What is the most direct acceptance criterion?
- Measure answer-bearing chunk recall and index-size inflation after splitting
- Check each quoted claim against the cited source record and revision
- Check indexed vector coverage and embedding failures by document batch
- Verify every returned chunk maps to the correct parent document and revision
Explanation
Correct Answer
C:The accepted change is to generate embeddings during indexing with a skillset; this evaluation checks whether it can avoid writing a custom service for every supported text-to-vector conversion, since integrated vectorization uses an indexer and enrichment pipeline to chunk and embed content; query-side vectorization still needs compatible settings.
Incorrect Answers
A:This test concerns the ability to keep retrieval passages within embedding and prompt size limits without losing section context, for which chunking creates smaller units for indexing and grounding; titles and parent identifiers should accompany chunks; it does not establish whether the change to generate embeddings during indexing with a skillset works for the current release.
B:This test concerns the ability to make an agent's grounded answer traceable back to the correct document chunk, for which a retrieval tool should return source identity and relevant snippet metadata along with content, not text alone; it does not establish whether the change to generate embeddings during indexing with a skillset works for the current release.
D:This test concerns the ability to search fine-grained passages while preserving links to complete source documents, for which index projections can place chunk records in a retrieval index with parent keys so answers can link back to source files; it does not establish whether the change to generate embeddings during indexing with a skillset works for the current release.
Question 471
After deployment, the team observes: A policy answer loses a threshold because the original value came from a table column header. Which remediation addresses the likely cause?
- Run layout-aware parsing and retain structural context in indexed passages
- Add a supported OCR or document-reading enrichment step before chunking
- Generate grounded image descriptions and connect them to the parent page and diagram reference
- Add appropriately typed filterable metadata fields and populate them during indexing
Explanation
Correct Answer
A:The fault signal is that a policy answer loses a threshold because the original value came from a table column header; the targeted fix works because layout processing detects document structure and table boundaries beyond simple plain-text OCR, restoring the ability to preserve relationships between headings, cells and surrounding text during document ingestion.
Incorrect Answers
B:If the failing function were to make text from photographed forms available for retrieval and citation, this change could be relevant because OCR converts visible characters into machine-readable text; text splitters cannot recover words from image-only pages; the actual symptom is that a policy answer loses a threshold because the original value came from a table column header, which calls for a different repair.
C:If the failing function were to index meaningful visual content of diagrams that contains no searchable text, this change could be relevant because image captioning or multimodal understanding produces descriptors for visual semantics; OCR alone reads only visible characters; the actual symptom is that a policy answer loses a threshold because the original value came from a table column header, which calls for a different repair.
D:If the failing function were to support both full-text answers and exact filtering by department and policy year, this change could be relevant because Azure AI Search index fields must be configured for the operations used; searchable text differs from structured filterable metadata; the actual symptom is that a policy answer loses a threshold because the original value came from a table column header, which calls for a different repair.
Question 472
Stakeholders want a Foundry-based application able to stop an agent from using search hits that the current requester is not allowed to read. Which choice implements that requirement?
- Filter retrieved evidence by user entitlements
- Define searchable and filterable fields appropriately
- Attach retrieved source identifiers to agent tools
- Apply retrieval quality evaluation before model tuning
Explanation
Correct Answer
A:For the stated requirement to stop an agent from using search hits that the current requester is not allowed to read, permission-aware retrieval or correctly enforced security filters keep restricted chunks out of prompt context; this is the mechanism that the system must implement.
Incorrect Answers
B:For a system whose objective is to support both full-text answers and exact filtering by department and policy year, Azure AI Search index fields must be configured for the operations used; searchable text differs from structured filterable metadata; this does not directly meet the required outcome to stop an agent from using search hits that the current requester is not allowed to read.
C:For a system whose objective is to make an agent's grounded answer traceable back to the correct document chunk, a retrieval tool should return source identity and relevant snippet metadata along with content, not text alone; this does not directly meet the required outcome to stop an agent from using search hits that the current requester is not allowed to read.
D:For a system whose objective is to diagnose whether poor RAG answers originate from missing passages rather than generation itself, search relevance and recall metrics isolate retrieval defects; changing prompts cannot create facts absent from the context; this does not directly meet the required outcome to stop an agent from using search hits that the current requester is not allowed to read.
Question 473
Engineers have implemented a change to filter retrieved evidence by user entitlements. What validation would expose a remaining defect?
- Compare indexer status and content versions with the source repository after updates
- Test normalized-field accuracy and failed-enrichment retry behavior
- Red-team searches with restricted accounts and audit unauthorized passage exposure
- Measure ingestion lag and stale-citation rate following a controlled source update
- Track retrieval recall at k separately from answer faithfulness and fluency
Explanation
Correct Answer
C:The accepted change is to filter retrieved evidence by user entitlements; this evaluation checks whether it can stop an agent from using search hits that the current requester is not allowed to read, since permission-aware retrieval or correctly enforced security filters keep restricted chunks out of prompt context.
Incorrect Answers
A:This test concerns the ability to keep search content synchronized with periodically changing files in a connected repository, for which Azure AI Search indexers read supported sources and drive enrichment and indexing on a schedule; external custom sources may need a push path; it does not establish whether the change to filter retrieved evidence by user entitlements works for the current release.
B:This test concerns the ability to apply a proprietary extraction algorithm that the built-in search skillset cannot express, for which custom Web API skills can call external enrichment code and return structured outputs to an indexer pipeline; it does not establish whether the change to filter retrieved evidence by user entitlements works for the current release.
D:This test concerns the ability to avoid answering from superseded policy versions after files change, for which scheduled or change-aware ingestion and indexed version checks keep the retrieval corpus aligned with its authority; it does not establish whether the change to filter retrieved evidence by user entitlements works for the current release.
E:This test concerns the ability to diagnose whether poor RAG answers originate from missing passages rather than generation itself, for which search relevance and recall metrics isolate retrieval defects; changing prompts cannot create facts absent from the context; it does not establish whether the change to filter retrieved evidence by user entitlements works for the current release.
Question 474
A design review proposes to match index vector dimensions to embedding model outputs. Which measurable check would establish success on real inputs?
- Compare top-k answer relevance and reranker scores on labeled questions
- Test filtered retrieval and inspect whether documents from other departments are excluded
- Index sample vectors, issue test queries, and verify zero dimension or vector-configuration errors
- Check each quoted claim against the cited source record and revision
Explanation
Correct Answer
C:The accepted change is to match index vector dimensions to embedding model outputs; this evaluation checks whether it can ensure the index's vector field accepts embeddings produced by the selected model, since an Azure AI Search vector field must have dimensions compatible with the embedding output and a valid vector search configuration; mismatched vectors fail at indexing or query time.
Incorrect Answers
A:This test concerns the ability to promote passages that directly answer questions over textually similar but less useful chunks, for which semantic ranking reranks a first-stage candidate list using language understanding; it cannot rescue documents never retrieved; it does not establish whether the change to match index vector dimensions to embedding model outputs works for the current release.
B:This test concerns the ability to support both full-text answers and exact filtering by department and policy year, for which Azure AI Search index fields must be configured for the operations used; searchable text differs from structured filterable metadata; it does not establish whether the change to match index vector dimensions to embedding model outputs works for the current release.
D:This test concerns the ability to make an agent's grounded answer traceable back to the correct document chunk, for which a retrieval tool should return source identity and relevant snippet metadata along with content, not text alone; it does not establish whether the change to match index vector dimensions to embedding model outputs works for the current release.
Question 475
After implementing a workflow to use an indexer for supported data sources, the team needs a robust release gate. Which check should it choose?
- Compare indexer status and content versions with the source repository after updates
- Test normalized-field accuracy and failed-enrichment retry behavior
- Check indexed vector coverage and embedding failures by document batch
- Verify every returned chunk maps to the correct parent document and revision
Explanation
Correct Answer
A:The accepted change is to use an indexer for supported data sources; this evaluation checks whether it can keep search content synchronized with periodically changing files in a connected repository, since Azure AI Search indexers read supported sources and drive enrichment and indexing on a schedule; external custom sources may need a push path.
Incorrect Answers
B:This test concerns the ability to apply a proprietary extraction algorithm that the built-in search skillset cannot express, for which custom Web API skills can call external enrichment code and return structured outputs to an indexer pipeline; it does not establish whether the change to use an indexer for supported data sources works for the current release.
C:This test concerns the ability to avoid writing a custom service for every supported text-to-vector conversion, for which integrated vectorization uses an indexer and enrichment pipeline to chunk and embed content; query-side vectorization still needs compatible settings; it does not establish whether the change to use an indexer for supported data sources works for the current release.
D:This test concerns the ability to search fine-grained passages while preserving links to complete source documents, for which index projections can place chunk records in a retrieval index with parent keys so answers can link back to source files; it does not establish whether the change to use an indexer for supported data sources works for the current release.
Question 476
A solution architect is deciding how to return invoice identifiers and dates in a fixed data contract from unstructured input. Which Azure AI pattern addresses that task?
- Generate concise contextual fields from source content
- Train or configure a custom extraction model for unusual forms
- Use source grounding and confidence for audit
- Configure an analyzer with extracted fields
Explanation
Correct Answer
D:For the stated requirement to return invoice identifiers and dates in a fixed data contract from unstructured input, a Content Understanding analyzer defines extraction fields and processing behavior and can emit structured results consistent with its schema; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to create an evidence-bounded description of a document section when no exact copied phrase suffices, generated analyzer fields derive structured descriptions from available source evidence; they should not be treated as direct verbatim extraction; this does not directly meet the required outcome to return invoice identifiers and dates in a fixed data contract from unstructured input.
B:For a system whose objective is to capture organization-specific fields from proprietary templates that prebuilt types do not recognize, a custom document extraction approach can target labeled form fields; deployment choice depends on samples and supported formats; this does not directly meet the required outcome to return invoice identifiers and dates in a fixed data contract from unstructured input.
C:For a system whose objective is to show the source region of each extracted critical field and send uncertain fields to review, document analyzer source grounding can attach field evidence and confidence estimates when enabled in supported configurations; this does not directly meet the required outcome to return invoice identifiers and dates in a fixed data contract from unstructured input.
Question 477
The support ticket describes this behavior: A finance workflow gets raw OCR text but employees manually locate totals on every standard invoice. Which engineering change should come first?
- Use the applicable prebuilt invoice extractor with field-level business checks
- Flag inconsistent or low-confidence fields for review before the posting transaction
- Run the supported OCR/read capability and store text with page references
- Apply layout analysis and use returned structural relationships in downstream parsing
Explanation
Correct Answer
A:The fault signal is that a finance workflow gets raw OCR text but employees manually locate totals on every standard invoice; the targeted fix works because a specialized prebuilt document model extracts familiar financial document fields and structure; it may still need validation for unfamiliar layouts, restoring the ability to capture supplier, invoice date, totals, and line items from ordinary invoices.
Incorrect Answers
B:If the failing function were to avoid automatically posting a financial record when source evidence is poor or conflicting, this change could be relevant because human-review queues use risk or confidence thresholds, with an auditable override process, to handle uncertain field extraction; the actual symptom is that a finance workflow gets raw OCR text but employees manually locate totals on every standard invoice, which calls for a different repair.
C:If the failing function were to extract machine-readable text from simple scanned letters that do not require typed business fields, this change could be relevant because document reading recovers words and lines; an invoice-specific schema adds overhead when plain transcription is the goal; the actual symptom is that a finance workflow gets raw OCR text but employees manually locate totals on every standard invoice, which calls for a different repair.
D:If the failing function were to extract document structure from forms containing tables, checkboxes and headings, this change could be relevant because the Document Intelligence layout model returns text plus structural elements such as tables and selection marks; the actual symptom is that a finance workflow gets raw OCR text but employees manually locate totals on every standard invoice, which calls for a different repair.
Question 478
The support ticket describes this behavior: A packet contains two contracts and one invoice, but the parser merges all dates into one record. Which engineering change should come first?
- Add document-type classification and route each category to the right processing branch
- Choose structure-preserving Markdown extraction and keep page or section metadata
- Enable appropriate segmentation and associate each section with its own extraction result
- Configure structured JSON fields and validate values before creating business records
Explanation
Correct Answer
C:The fault signal is that a packet contains two contracts and one invoice, but the parser merges all dates into one record; the targeted fix works because Content Understanding segmentation separates heterogeneous portions for targeted processing; treating the entire file as one document can conflate values, restoring the ability to split a combined upload into logical subdocuments before classifying and extracting their fields.
Incorrect Answers
A:If the failing function were to decide whether an upload is an insurance claim, receipt, or contract before specialized extraction, this change could be relevant because classify fields map content to a defined category set, enabling downstream routing to appropriate logic; the actual symptom is that a packet contains two contracts and one invoice, but the parser merges all dates into one record, which calls for a different repair.
B:If the failing function were to prepare a complex report for RAG ingestion with headings and tabular context intact, this change could be relevant because structured Markdown can preserve readable hierarchy for chunking and retrieval better than flattened strings; the actual symptom is that a packet contains two contracts and one invoice, but the parser merges all dates into one record, which calls for a different repair.
D:If the failing function were to populate a line-of-business workflow with predictable field names, types, and nested items, this change could be relevant because schema-directed JSON output supports deterministic parsing and downstream validation; a human-readable summary is not a reliable API contract; the actual symptom is that a packet contains two contracts and one invoice, but the parser merges all dates into one record, which calls for a different repair.
Question 479
After implementing a workflow to return JSON matching an automation schema, the team needs a robust release gate. Which check should it choose?
- Verify source regions and reviewer correction rates across confidence bands
- Measure incorrect automatic posting rate and reviewer resolution time on ambiguous scans
- Test heading-to-paragraph alignment and retrieval of table-dependent answers
- Measure parse success and semantic field accuracy across varied document layouts
Explanation
Correct Answer
D:The accepted change is to return JSON matching an automation schema; this evaluation checks whether it can populate a line-of-business workflow with predictable field names, types, and nested items, since schema-directed JSON output supports deterministic parsing and downstream validation; a human-readable summary is not a reliable API contract.
Incorrect Answers
A:This test concerns the ability to show the source region of each extracted critical field and send uncertain fields to review, for which document analyzer source grounding can attach field evidence and confidence estimates when enabled in supported configurations; it does not establish whether the change to return JSON matching an automation schema works for the current release.
B:This test concerns the ability to avoid automatically posting a financial record when source evidence is poor or conflicting, for which human-review queues use risk or confidence thresholds, with an auditable override process, to handle uncertain field extraction; it does not establish whether the change to return JSON matching an automation schema works for the current release.
C:This test concerns the ability to prepare a complex report for RAG ingestion with headings and tabular context intact, for which structured Markdown can preserve readable hierarchy for chunking and retrieval better than flattened strings; it does not establish whether the change to return JSON matching an automation schema works for the current release.
Question 480
A new customer-facing AI service must show the source region of each extracted critical field and send uncertain fields to review. Which approach should the team select first?
- Generate concise contextual fields from source content
- Use classification fields to route document types
- Use a read or OCR model for page text
- Use source grounding and confidence for audit
Explanation
Correct Answer
D:For the stated requirement to show the source region of each extracted critical field and send uncertain fields to review, document analyzer source grounding can attach field evidence and confidence estimates when enabled in supported configurations; this is the mechanism that the system must implement.
Incorrect Answers
A:For a system whose objective is to create an evidence-bounded description of a document section when no exact copied phrase suffices, generated analyzer fields derive structured descriptions from available source evidence; they should not be treated as direct verbatim extraction; this does not directly meet the required outcome to show the source region of each extracted critical field and send uncertain fields to review.
B:For a system whose objective is to decide whether an upload is an insurance claim, receipt, or contract before specialized extraction, classify fields map content to a defined category set, enabling downstream routing to appropriate logic; this does not directly meet the required outcome to show the source region of each extracted critical field and send uncertain fields to review.
C:For a system whose objective is to extract machine-readable text from simple scanned letters that do not require typed business fields, document reading recovers words and lines; an invoice-specific schema adds overhead when plain transcription is the goal; this does not directly meet the required outcome to show the source region of each extracted critical field and send uncertain fields to review.