McAfee-Secured Website

Exam Code: AI-300

Exam Name: Operationalizing Machine Learning and Generative AI Solutions

Certification Provider: Microsoft

Corresponding Certification: Microsoft Certified: Machine Learning Operations Engineer Associate

100% Microsoft AI-300 Guaranteed Success With Testking Training Materials

Get Real AI-300 Exam Questions & Expert Verified Answers as Experienced on the Real Exam!

213 Questions & Answers with Testing Engine

"Operationalizing Machine Learning and Generative AI Solutions Exam", also known as AI-300 exam, is a Microsoft certification exam.

Pass your tests with the always up-to-date AI-300 Exam Engine. Your AI-300 training materials keep you at the head of the pack!

Satisfaction Guaranteed

Money Back Guarantee

Testking provides hassle-free money back guarantee with our products. That is because we have 100% trust in the abilities of our professional and experience product team, and our record is a proof of that.

99.6% PASS RATE
Was: $137.49
Now: $124.99

Microsoft AI-300 Practice Test 70

Question 690

A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. How can the index preserve this cross-boundary context?

  1. Evaluate stricter relevance thresholds on a held-out question and document relevance set.
  2. Increase candidate top-k and measure recall before imposing a bounded context budget.
  3. Add controlled overlap or parent-section context around boundary-sensitive chunks.
  4. Apply enforced tenant and authorization filters in retrieval before documents enter model context.

Explanation

Correct Answer

C:

Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units.

Incorrect Answers

A:

Evaluate stricter relevance thresholds on a held-out question and document relevance set. It addresses another case: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. For that case, the distinguishing justification or acceptance check is: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries. The present case instead states: A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. It calls for Add controlled overlap or parent-section context around boundary-sensitive chunks. Showing success for the other case would not establish: Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units.

B:

The alternative, Increase candidate top-k and measure recall before imposing a bounded context budget, belongs to a different technical condition: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. There the source-specific mechanism or proof is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed. The condition asked about here is: A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. A suitable correction is Add controlled overlap or parent-section context around boundary-sensitive chunks; the targeted justification or result is: Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units.

D:

Apply enforced tenant and authorization filters in retrieval before documents enter model context. That operation would be relevant if: A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. Its corresponding reason or verification is: Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read. However the problem in this question is: A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. The requirement instead calls for Add controlled overlap or parent-section context around boundary-sensitive chunks; the question-specific check is: Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units.

Question 691

A vector query returns zero references after a developer imposes a fixed minimum score of 0.95 copied from another index. What should be done?

  1. Decompose the compound request into focused retrieval queries and merge supported evidence.
  2. Calibrate the cutoff using relevant labeled queries and the actual score scale of this retrieval configuration.
  3. Filter or rank by verified effective-version metadata while excluding superseded document revisions.
  4. Add controlled overlap or parent-section context around boundary-sensitive chunks.

Explanation

Correct Answer

B:

Similarity and fused retrieval scores are not interchangeable; a borrowed cutoff can wrongly discard useful results.

Incorrect Answers

A:

Decompose the compound request into focused retrieval queries and merge supported evidence. This is useful for the distinct issue: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. The relevant technical rationale or evidence for that issue is: Targeted subqueries cover independent constraints that one broad embedding may not rank together. Here the reported issue is: A vector query returns zero references after a developer imposes a fixed minimum score of 0.95 copied from another index. The relevant response is: Calibrate the cutoff using relevant labeled queries and the actual score scale of this retrieval configuration. The first result cannot verify this different correction: Similarity and fused retrieval scores are not interchangeable; a borrowed cutoff can wrongly discard useful results.

C:

Filter or rank by verified effective-version metadata while excluding superseded document revisions. It addresses another case: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. For that case, the distinguishing justification or acceptance check is: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance. The present case instead states: A vector query returns zero references after a developer imposes a fixed minimum score of 0.95 copied from another index. It calls for Calibrate the cutoff using relevant labeled queries and the actual score scale of this retrieval configuration. Showing success for the other case would not establish: Similarity and fused retrieval scores are not interchangeable; a borrowed cutoff can wrongly discard useful results.

D:

The alternative, Add controlled overlap or parent-section context around boundary-sensitive chunks, belongs to a different technical condition: A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. There the source-specific mechanism or proof is: Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units. The condition asked about here is: A vector query returns zero references after a developer imposes a fixed minimum score of 0.95 copied from another index. A suitable correction is Calibrate the cutoff using relevant labeled queries and the actual score scale of this retrieval configuration; the targeted justification or result is: Similarity and fused retrieval scores are not interchangeable; a borrowed cutoff can wrongly discard useful results.

Question 692

A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. Which corrective experiment is most relevant?

  1. Evaluate stricter relevance thresholds on a held-out question and document relevance set.
  2. Apply enforced tenant and authorization filters in retrieval before documents enter model context.
  3. Increase candidate top-k and measure recall before imposing a bounded context budget.
  4. Filter or rank by verified effective-version metadata while excluding superseded document revisions.

Explanation

Correct Answer

A:

Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries.

Incorrect Answers

B:

Apply enforced tenant and authorization filters in retrieval before documents enter model context. A separate source case supports that approach when: A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. The supporting rationale or observable result is: Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read. This question describes a different issue: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. Address it through Evaluate stricter relevance thresholds on a held-out question and document relevance set, with the required justification or evidence: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries.

C:

Increase candidate top-k and measure recall before imposing a bounded context budget. This is useful for the distinct issue: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. The relevant technical rationale or evidence for that issue is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed. Here the reported issue is: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. The relevant response is: Evaluate stricter relevance thresholds on a held-out question and document relevance set. The first result cannot verify this different correction: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries.

D:

Filter or rank by verified effective-version metadata while excluding superseded document revisions. It addresses another case: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. For that case, the distinguishing justification or acceptance check is: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance. The present case instead states: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. It calls for Evaluate stricter relevance thresholds on a held-out question and document relevance set. Showing success for the other case would not establish: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries.

Question 693

A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. What tuning trial is appropriate?

  1. Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas.
  2. Verify ingestion change detection and refresh the index entries for the revised document version.
  3. Split source manuals into coherent section-level chunks with headings and measured overlap.
  4. Increase candidate top-k and measure recall before imposing a bounded context budget.

Explanation

Correct Answer

D:

A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed.

Incorrect Answers

A:

Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas. That operation would be relevant if: A hybrid search deployment starts returning 429 errors after engineers increase search candidate count and vector graph effort sharply. Its corresponding reason or verification is: Excessive retrieval work can cause throttling; workload tuning should precede infrastructure scale-out. However the problem in this question is: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. The requirement instead calls for Increase candidate top-k and measure recall before imposing a bounded context budget; the question-specific check is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed.

B:

Verify ingestion change detection and refresh the index entries for the revised document version. A separate source case supports that approach when: A knowledge page was corrected yesterday but retrieved excerpts still contain the deleted error. The supporting rationale or observable result is: Retrieval cannot reflect corrected knowledge until obsolete chunks and embeddings have been replaced or invalidated. This question describes a different issue: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. Address it through Increase candidate top-k and measure recall before imposing a bounded context budget, with the required justification or evidence: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed.

C:

Split source manuals into coherent section-level chunks with headings and measured overlap. This is useful for the distinct issue: A RAG assistant misses policy exceptions because the ingestion job stores 18-page manuals as single vectors. The relevant technical rationale or evidence for that issue is: Smaller coherent sections surface relevant exceptions without embedding unrelated manual chapters in the same vector. Here the reported issue is: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. The relevant response is: Increase candidate top-k and measure recall before imposing a bounded context budget. The first result cannot verify this different correction: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed.

Question 694

A retriever sends forty full documents to a small-context model, causing truncation of the most useful passage. What should be optimized?

  1. Test a multilingual embedding and language-aware query processing on labeled cross-language questions.
  2. Split source manuals into coherent section-level chunks with headings and measured overlap.
  3. Rerank candidates and include only high-value excerpts that fit the generation context window.
  4. Verify ingestion change detection and refresh the index entries for the revised document version.

Explanation

Correct Answer

C:

Selective context assembly preserves relevant evidence without overwhelming the model with low-ranked material.

Incorrect Answers

A:

The alternative, Test a multilingual embedding and language-aware query processing on labeled cross-language questions, belongs to a different technical condition: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. There the source-specific mechanism or proof is: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores. The condition asked about here is: A retriever sends forty full documents to a small-context model, causing truncation of the most useful passage. A suitable correction is Rerank candidates and include only high-value excerpts that fit the generation context window; the targeted justification or result is: Selective context assembly preserves relevant evidence without overwhelming the model with low-ranked material.

B:

Split source manuals into coherent section-level chunks with headings and measured overlap. That operation would be relevant if: A RAG assistant misses policy exceptions because the ingestion job stores 18-page manuals as single vectors. Its corresponding reason or verification is: Smaller coherent sections surface relevant exceptions without embedding unrelated manual chapters in the same vector. However the problem in this question is: A retriever sends forty full documents to a small-context model, causing truncation of the most useful passage. The requirement instead calls for Rerank candidates and include only high-value excerpts that fit the generation context window; the question-specific check is: Selective context assembly preserves relevant evidence without overwhelming the model with low-ranked material.

D:

Verify ingestion change detection and refresh the index entries for the revised document version. A separate source case supports that approach when: A knowledge page was corrected yesterday but retrieved excerpts still contain the deleted error. The supporting rationale or observable result is: Retrieval cannot reflect corrected knowledge until obsolete chunks and embeddings have been replaced or invalidated. This question describes a different issue: A retriever sends forty full documents to a small-context model, causing truncation of the most useful passage. Address it through Rerank candidates and include only high-value excerpts that fit the generation context window, with the required justification or evidence: Selective context assembly preserves relevant evidence without overwhelming the model with low-ranked material.

Question 695

A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. What must change?

  1. Evaluate stricter relevance thresholds on a held-out question and document relevance set.
  2. Apply enforced tenant and authorization filters in retrieval before documents enter model context.
  3. Use layout-aware table extraction and keep associated headers with their data rows.
  4. Test a multilingual embedding and language-aware query processing on labeled cross-language questions.

Explanation

Correct Answer

B:

Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read.

Incorrect Answers

A:

Evaluate stricter relevance thresholds on a held-out question and document relevance set. It addresses another case: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. For that case, the distinguishing justification or acceptance check is: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries. The present case instead states: A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. It calls for Apply enforced tenant and authorization filters in retrieval before documents enter model context. Showing success for the other case would not establish: Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read.

C:

The alternative, Use layout-aware table extraction and keep associated headers with their data rows, belongs to a different technical condition: Financial statements are chunked in the middle of a multi-row table, separating column labels from values. There the source-specific mechanism or proof is: Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them. The condition asked about here is: A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. A suitable correction is Apply enforced tenant and authorization filters in retrieval before documents enter model context; the targeted justification or result is: Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read.

D:

Test a multilingual embedding and language-aware query processing on labeled cross-language questions. That operation would be relevant if: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. Its corresponding reason or verification is: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores. However the problem in this question is: A tenant-specific RAG search sometimes returns documents belonging to another customer. The model is instructed to ignore them, but they remain in context. The requirement instead calls for Apply enforced tenant and authorization filters in retrieval before documents enter model context; the question-specific check is: Prompt instructions cannot replace access-controlled retrieval of only documents the requester is entitled to read.

Question 696

The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. How should retrieval be constrained?

  1. Filter or rank by verified effective-version metadata while excluding superseded document revisions.
  2. Increase candidate top-k and measure recall before imposing a bounded context budget.
  3. Rerank candidates and include only high-value excerpts that fit the generation context window.
  4. Decompose the compound request into focused retrieval queries and merge supported evidence.

Explanation

Correct Answer

A:

Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance.

Incorrect Answers

B:

Increase candidate top-k and measure recall before imposing a bounded context budget. This is useful for the distinct issue: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. The relevant technical rationale or evidence for that issue is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed. Here the reported issue is: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. The relevant response is: Filter or rank by verified effective-version metadata while excluding superseded document revisions. The first result cannot verify this different correction: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance.

C:

Rerank candidates and include only high-value excerpts that fit the generation context window. It addresses another case: A retriever sends forty full documents to a small-context model, causing truncation of the most useful passage. For that case, the distinguishing justification or acceptance check is: Selective context assembly preserves relevant evidence without overwhelming the model with low-ranked material. The present case instead states: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. It calls for Filter or rank by verified effective-version metadata while excluding superseded document revisions. Showing success for the other case would not establish: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance.

D:

The alternative, Decompose the compound request into focused retrieval queries and merge supported evidence, belongs to a different technical condition: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. There the source-specific mechanism or proof is: Targeted subqueries cover independent constraints that one broad embedding may not rank together. The condition asked about here is: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. A suitable correction is Filter or rank by verified effective-version metadata while excluding superseded document revisions; the targeted justification or result is: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance.

Question 697

A hybrid search deployment starts returning 429 errors after engineers increase search candidate count and vector graph effort sharply. Which first correction fits?

  1. Test a multilingual embedding and language-aware query processing on labeled cross-language questions.
  2. Verify ingestion change detection and refresh the index entries for the revised document version.
  3. Increase candidate top-k and measure recall before imposing a bounded context budget.
  4. Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas.

Explanation

Correct Answer

D:

Excessive retrieval work can cause throttling; workload tuning should precede infrastructure scale-out.

Incorrect Answers

A:

Test a multilingual embedding and language-aware query processing on labeled cross-language questions. A separate source case supports that approach when: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. The supporting rationale or observable result is: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores. This question describes a different issue: A hybrid search deployment starts returning 429 errors after engineers increase search candidate count and vector graph effort sharply. Address it through Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas, with the required justification or evidence: Excessive retrieval work can cause throttling; workload tuning should precede infrastructure scale-out.

B:

Verify ingestion change detection and refresh the index entries for the revised document version. This is useful for the distinct issue: A knowledge page was corrected yesterday but retrieved excerpts still contain the deleted error. The relevant technical rationale or evidence for that issue is: Retrieval cannot reflect corrected knowledge until obsolete chunks and embeddings have been replaced or invalidated. Here the reported issue is: A hybrid search deployment starts returning 429 errors after engineers increase search candidate count and vector graph effort sharply. The relevant response is: Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas. The first result cannot verify this different correction: Excessive retrieval work can cause throttling; workload tuning should precede infrastructure scale-out.

C:

Increase candidate top-k and measure recall before imposing a bounded context budget. It addresses another case: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. For that case, the distinguishing justification or acceptance check is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed. The present case instead states: A hybrid search deployment starts returning 429 errors after engineers increase search candidate count and vector graph effort sharply. It calls for Reduce expensive search parameters and candidate volume, then retest latency before adding search replicas. Showing success for the other case would not establish: Excessive retrieval work can cause throttling; workload tuning should precede infrastructure scale-out.

Question 698

Financial statements are chunked in the middle of a multi-row table, separating column labels from values. Which preprocessing strategy helps?

  1. Add controlled overlap or parent-section context around boundary-sensitive chunks.
  2. Decompose the compound request into focused retrieval queries and merge supported evidence.
  3. Use layout-aware table extraction and keep associated headers with their data rows.
  4. Return a supported no-answer or escalation response instead of forcing an unsupported citation.

Explanation

Correct Answer

C:

Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them.

Incorrect Answers

A:

Add controlled overlap or parent-section context around boundary-sensitive chunks. That operation would be relevant if: A benefits answer cites a chunk that ends just before the sentence containing the required exclusion. Its corresponding reason or verification is: Boundary context keeps the qualification beside the rule instead of losing it in adjacent retrieval units. However the problem in this question is: Financial statements are chunked in the middle of a multi-row table, separating column labels from values. The requirement instead calls for Use layout-aware table extraction and keep associated headers with their data rows; the question-specific check is: Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them.

B:

Decompose the compound request into focused retrieval queries and merge supported evidence. A separate source case supports that approach when: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. The supporting rationale or observable result is: Targeted subqueries cover independent constraints that one broad embedding may not rank together. This question describes a different issue: Financial statements are chunked in the middle of a multi-row table, separating column labels from values. Address it through Use layout-aware table extraction and keep associated headers with their data rows, with the required justification or evidence: Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them.

D:

Return a supported no-answer or escalation response instead of forcing an unsupported citation. This is useful for the distinct issue: A high-risk HR assistant receives no sufficiently relevant passage for a leave-policy question. The relevant technical rationale or evidence for that issue is: An explicit evidence threshold should prevent fabricated answers when retrieval cannot supply adequate source support. Here the reported issue is: Financial statements are chunked in the middle of a multi-row table, separating column labels from values. The relevant response is: Use layout-aware table extraction and keep associated headers with their data rows. The first result cannot verify this different correction: Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them.

Question 699

A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. What approach improves recall?

  1. Return a supported no-answer or escalation response instead of forcing an unsupported citation.
  2. Decompose the compound request into focused retrieval queries and merge supported evidence.
  3. Filter or rank by verified effective-version metadata while excluding superseded document revisions.
  4. Evaluate stricter relevance thresholds on a held-out question and document relevance set.

Explanation

Correct Answer

B:

Targeted subqueries cover independent constraints that one broad embedding may not rank together.

Incorrect Answers

A:

The alternative, Return a supported no-answer or escalation response instead of forcing an unsupported citation, belongs to a different technical condition: A high-risk HR assistant receives no sufficiently relevant passage for a leave-policy question. There the source-specific mechanism or proof is: An explicit evidence threshold should prevent fabricated answers when retrieval cannot supply adequate source support. The condition asked about here is: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. A suitable correction is Decompose the compound request into focused retrieval queries and merge supported evidence; the targeted justification or result is: Targeted subqueries cover independent constraints that one broad embedding may not rank together.

C:

Filter or rank by verified effective-version metadata while excluding superseded document revisions. That operation would be relevant if: The assistant answers a 2026 policy query using a superseded 2023 procedure stored beside the current version. Its corresponding reason or verification is: Version-aware retrieval prevents obsolete but semantically similar policy language from displacing current approved guidance. However the problem in this question is: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. The requirement instead calls for Decompose the compound request into focused retrieval queries and merge supported evidence; the question-specific check is: Targeted subqueries cover independent constraints that one broad embedding may not rank together.

D:

Evaluate stricter relevance thresholds on a held-out question and document relevance set. A separate source case supports that approach when: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. The supporting rationale or observable result is: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries. This question describes a different issue: A customer asks for a comparison requiring a policy rule and a separate exception memo; one broad search misses the memo. Address it through Decompose the compound request into focused retrieval queries and merge supported evidence, with the required justification or evidence: Targeted subqueries cover independent constraints that one broad embedding may not rank together.

Question 700

A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. What is the next retrieval experiment?

  1. Test a multilingual embedding and language-aware query processing on labeled cross-language questions.
  2. Use layout-aware table extraction and keep associated headers with their data rows.
  3. Increase candidate top-k and measure recall before imposing a bounded context budget.
  4. Evaluate stricter relevance thresholds on a held-out question and document relevance set.

Explanation

Correct Answer

A:

Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores.

Incorrect Answers

B:

Use layout-aware table extraction and keep associated headers with their data rows. That operation would be relevant if: Financial statements are chunked in the middle of a multi-row table, separating column labels from values. Its corresponding reason or verification is: Preserving table structure prevents retrieval of orphaned numeric values without the column labels needed to interpret them. However the problem in this question is: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. The requirement instead calls for Test a multilingual embedding and language-aware query processing on labeled cross-language questions; the question-specific check is: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores.

C:

Increase candidate top-k and measure recall before imposing a bounded context budget. A separate source case supports that approach when: A product-support bot retrieves only one chunk per query and misses second-step configuration details. The index contains both steps. The supporting rationale or observable result is: A larger candidate pool can expose missing relevant steps, but quality and latency must be measured rather than assumed. This question describes a different issue: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. Address it through Test a multilingual embedding and language-aware query processing on labeled cross-language questions, with the required justification or evidence: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores.

D:

Evaluate stricter relevance thresholds on a held-out question and document relevance set. This is useful for the distinct issue: A legal RAG assistant cites unrelated contracts because similarity filtering is too permissive. The relevant technical rationale or evidence for that issue is: Thresholds should reject known irrelevant chunks while preserving relevant contract clauses under representative queries. Here the reported issue is: A multilingual knowledge base retrieves English summaries poorly when requests are in Spanish and French. The relevant response is: Test a multilingual embedding and language-aware query processing on labeled cross-language questions. The first result cannot verify this different correction: Cross-language relevance should be measured using actual multilingual queries, not assumed from English-only retrieval scores.

Operationalizing Machine Learning and Generative AI Solutions (AI-300) Practice Tests