Question 150
A RAG model receives two conflicting policy documents and chooses a rescinded version in its response. Which TWO corrections make the evidence more reliable? (Choose TWO.)
- Use structure-aware boundaries and suitable overlap between adjacent chunks
- Run embedding generation for the new content before expecting vector-neighbor retrieval
- State explicit source-priority and attribution rules when building the augmented prompt
- Construct citations from actual retrieved source metadata rather than free-form model guesses
- Filter the retrieved passages to authoritative currently effective documents
Explanation
Correct Answers
C:Source precedence and attribution help the model distinguish authoritative policy evidence from superseded or informal notes.
E:Grounding on valid source versions prevents the model from using stale material merely because it is semantically close to the question.
Incorrect Answers
A:Meaningful boundaries preserve local context; blindly fixed-length splitting can separate definitions from their qualifiers in its own workflow, but that is a different correction from the present requirement to filter the retrieved passages to authoritative currently effective documents and state explicit source-priority and attribution rules when building the augmented prompt. The reported issue would remain.
B:That change addresses another failure mode: text fields alone do not create embeddings unless supported vectorization is configured; it does not address this case, for which the relevant intervention is to filter the retrieved passages to authoritative currently effective documents and state explicit source-priority and attribution rules when building the augmented prompt.
D:Although useful elsewhere in the private-document answer assistant, source metadata is application evidence; unconstrained text generation can fabricate bibliographic details; the present case is different and calls for the team to filter the retrieved passages to authoritative currently effective documents and state explicit source-priority and attribution rules when building the augmented prompt.
Question 151
A private-document answer assistant must encode each passage as a numeric vector for semantic similarity. Reviewers will accept the implementation when each chunk has a vector representation. Which approach best fulfills that requirement?
- Chunk source documents into passages
- Attach source metadata and citations to answers
- Retrieve relevant passages from a search index
- Generate embeddings for document chunks
- Augment the model prompt with retrieved evidence
Explanation
Correct Answer
D:An embedding model maps text into a vector space where related meaning can be compared mathematically. For this private-document answer assistant, the intended evidence is that each chunk has a vector representation. Embeddings are not natural-language answers and do not enforce access to private documents.
Incorrect Answers
A:Chunking produces passages sized for retrieval and model context; metadata can preserve document lineage; it may be used to split lengthy source documents into retrievable sections with appropriate overlap in a private-document answer assistant, but the current requirement is instead to encode each passage as a numeric vector for semantic similarity. That separate capability does not prove that each chunk has a vector representation; It does not by itself convert passages into semantic vectors or compose final answers, so selecting that alternative would not meet the stated acceptance check.
B:Citation-aware output carries source identifiers, titles, or links forward from retrieval into the response; it may be used to let a user trace an AI answer back to the retrieved documents used as support in a private-document answer assistant, but the current requirement is instead to encode each passage as a numeric vector for semantic similarity. That separate capability does not prove that each chunk has a vector representation; Citations are not created merely by computing embeddings if source metadata is discarded, so selecting that alternative would not meet the stated acceptance check.
C:Retrieval queries a prepared index to select likely relevant documents using keyword, vector, or hybrid techniques; it may be used to find supporting content for a user's current question before generation in a private-document answer assistant, but the current requirement is instead to encode each passage as a numeric vector for semantic similarity. That separate capability does not prove that each chunk has a vector representation; It does not guarantee the chat model will cite evidence unless the retrieved material is passed along and used, so selecting that alternative would not meet the stated acceptance check.
E:The augmentation step puts retrieved context into the model input so generation can use current enterprise data; it may be used to provide the selected source passages alongside the question and instructions in a private-document answer assistant, but the current requirement is instead to encode each passage as a numeric vector for semantic similarity. That separate capability does not prove that each chunk has a vector representation; It does not update the model's pretrained weights or replace the need for retrieval, so selecting that alternative would not meet the stated acceptance check.
Question 152
The solution brief for a private-document answer assistant requires the system to find supporting content for a user's current question before generation. A successful test confirms that the request identifies a relevant grounded-content set. What should the team use?
- Augment the model prompt with retrieved evidence
- Retrieve relevant passages from a search index
- Attach source metadata and citations to answers
- Chunk source documents into passages
Explanation
Correct Answer
B:Retrieval queries a prepared index to select likely relevant documents using keyword, vector, or hybrid techniques. For this private-document answer assistant, the intended evidence is that the request identifies a relevant grounded-content set. It does not guarantee the chat model will cite evidence unless the retrieved material is passed along and used.
Incorrect Answers
A:The augmentation step puts retrieved context into the model input so generation can use current enterprise data; it may be used to provide the selected source passages alongside the question and instructions in a private-document answer assistant, but the current requirement is instead to find supporting content for a user's current question before generation. That separate capability does not prove that the request identifies a relevant grounded-content set; It does not update the model's pretrained weights or replace the need for retrieval, so selecting that alternative would not meet the stated acceptance check.
C:Citation-aware output carries source identifiers, titles, or links forward from retrieval into the response; it may be used to let a user trace an AI answer back to the retrieved documents used as support in a private-document answer assistant, but the current requirement is instead to find supporting content for a user's current question before generation. That separate capability does not prove that the request identifies a relevant grounded-content set; Citations are not created merely by computing embeddings if source metadata is discarded, so selecting that alternative would not meet the stated acceptance check.
D:Chunking produces passages sized for retrieval and model context; metadata can preserve document lineage; it may be used to split lengthy source documents into retrievable sections with appropriate overlap in a private-document answer assistant, but the current requirement is instead to find supporting content for a user's current question before generation. That separate capability does not prove that the request identifies a relevant grounded-content set; It does not by itself convert passages into semantic vectors or compose final answers, so selecting that alternative would not meet the stated acceptance check.
Question 153
Search results repeatedly return the first page of a long policy, missing a later exception clause. What ingestion adjustment would help?
- Split the policy into separately retrievable passages with source-page metadata
- Inject the selected evidence passages into the model request context
- Keep stable document IDs and page or chunk locators in retrieval results
- Apply supported metadata filters such as tenant and document validity before grounding
Explanation
Correct Answer
A:Chunked retrieval makes later portions independently searchable rather than representing an entire long file as one item. In the private-document answer assistant, confirm the fix by retesting the affected behavior after splitting the policy into separately retrievable passages with source-page metadata.
Incorrect Answers
B:Although useful elsewhere in the private-document answer assistant, retrieval does not ground generation until the passages are actually supplied to the model's input; the present case is different and calls for the team to split the policy into separately retrievable passages with source-page metadata.
C:Fine-grained source locators make generated citations navigable and auditable in its own workflow, but that is a different correction from the present requirement to split the policy into separately retrievable passages with source-page metadata. The reported issue would remain.
D:That change addresses another failure mode: filtering the candidate corpus by structured constraints reduces irrelevant or unauthorized retrieval results; it does not address this case, for which the relevant intervention is to split the policy into separately retrievable passages with source-page metadata.
Question 154
A generator sees the question but no retrieved documents even though search returned good matches. Which integration step is missing?
- Keep stable document IDs and page or chunk locators in retrieval results
- Use compatible embedding models and dimensions for indexed documents and query inputs
- Inject the selected evidence passages into the model request context
- Apply supported metadata filters such as tenant and document validity before grounding
Explanation
Correct Answer
C:Retrieval does not ground generation until the passages are actually supplied to the model's input. In the private-document answer assistant, confirm the fix by retesting the affected behavior after injecting the selected evidence passages into the model request context.
Incorrect Answers
A:That change addresses another failure mode: fine-grained source locators make generated citations navigable and auditable; it does not address this case, for which the relevant intervention is to inject the selected evidence passages into the model request context.
B:Although useful elsewhere in the private-document answer assistant, vector comparison requires schema-compatible vectors in the same intended representation space; the present case is different and calls for the team to inject the selected evidence passages into the model request context.
D:Filtering the candidate corpus by structured constraints reduces irrelevant or unauthorized retrieval results in its own workflow, but that is a different correction from the present requirement to inject the selected evidence passages into the model request context. The reported issue would remain.
Question 155
A team proposes treating the cosine similarity score as a calibrated factual-truth probability. What should be explained?
- Check indexed searchable fields, query formulation, and retrieval configuration
- Interpret similarity as retrieval relevance rather than a guarantee that a passage is correct
- Require a supported-answer fallback when the grounding set is empty or insufficient
- Record source version or retrieval timestamp alongside the cited document identifier
Explanation
Correct Answer
B:Embedding proximity estimates semantic relatedness, not the authenticity or correctness of a source claim. In the private-document answer assistant, confirm the fix by retesting the affected behavior after interpreting similarity as retrieval relevance rather than a guarantee that a passage is correct.
Incorrect Answers
A:Content cannot be retrieved when it was not indexed into searchable fields or the request addresses the wrong index/field in its own workflow, but that is a different correction from the present requirement to interpret similarity as retrieval relevance rather than a guarantee that a passage is correct. The reported issue would remain.
C:That change addresses another failure mode: rAG should not manufacture facts in the absence of evidence; an explicit abstention path is safer; it does not address this case, for which the relevant intervention is to interpret similarity as retrieval relevance rather than a guarantee that a passage is correct.
D:Although useful elsewhere in the private-document answer assistant, versioned provenance helps determine which source text supported an earlier answer; the present case is different and calls for the team to interpret similarity as retrieval relevance rather than a guarantee that a passage is correct.
Question 156
Users click a citation and reach the wrong page of a long handbook. What metadata should ingestion retain?
- Apply supported metadata filters such as tenant and document validity before grounding
- Inject the selected evidence passages into the model request context
- Keep stable document IDs and page or chunk locators in retrieval results
- Split the policy into separately retrievable passages with source-page metadata
Explanation
Correct Answer
C:Fine-grained source locators make generated citations navigable and auditable. In the private-document answer assistant, confirm the fix by retesting the affected behavior after keeping stable document IDs and page or chunk locators in retrieval results.
Incorrect Answers
A:Filtering the candidate corpus by structured constraints reduces irrelevant or unauthorized retrieval results in its own workflow, but that is a different correction from the present requirement to keep stable document IDs and page or chunk locators in retrieval results. The reported issue would remain.
B:That change addresses another failure mode: retrieval does not ground generation until the passages are actually supplied to the model's input; it does not address this case, for which the relevant intervention is to keep stable document IDs and page or chunk locators in retrieval results.
D:Although useful elsewhere in the private-document answer assistant, chunked retrieval makes later portions independently searchable rather than representing an entire long file as one item; the present case is different and calls for the team to keep stable document IDs and page or chunk locators in retrieval results.
Question 157
Two similar policies have different effective dates. What should the retriever enforce before choosing grounding text?
- Run embedding generation for the new content before expecting vector-neighbor retrieval
- Construct citations from actual retrieved source metadata rather than free-form model guesses
- Filter to the currently effective, access-authorized policy version
- Pass authoritative current sources with clear attribution and precedence instructions
Explanation
Correct Answer
C:A semantic match can still be stale or out of scope; retrieval must honor business applicability. In the private-document answer assistant, confirm the fix by retesting the affected behavior after filtering to the currently effective, access-authorized policy version.
Incorrect Answers
A:That change addresses another failure mode: text fields alone do not create embeddings unless supported vectorization is configured; it does not address this case, for which the relevant intervention is to filter to the currently effective, access-authorized policy version.
B:Source metadata is application evidence; unconstrained text generation can fabricate bibliographic details in its own workflow, but that is a different correction from the present requirement to filter to the currently effective, access-authorized policy version. The reported issue would remain.
D:Although useful elsewhere in the private-document answer assistant, a pile of unqualified passages can encourage unsupported synthesis; grounding should make source authority and context explicit; the present case is different and calls for the team to filter to the currently effective, access-authorized policy version.
Question 158
A chunk splitter cuts technical definitions mid-sentence and the model misinterprets them. What setting should be reconsidered?
- Pass authoritative current sources with clear attribution and precedence instructions
- Construct citations from actual retrieved source metadata rather than free-form model guesses
- Use structure-aware boundaries and suitable overlap between adjacent chunks
- Filter to the currently effective, access-authorized policy version
Explanation
Correct Answer
C:Meaningful boundaries preserve local context; blindly fixed-length splitting can separate definitions from their qualifiers. In the private-document answer assistant, confirm the fix by retesting the affected behavior after using structure-aware boundaries and suitable overlap between adjacent chunks.
Incorrect Answers
A:Although useful elsewhere in the private-document answer assistant, a pile of unqualified passages can encourage unsupported synthesis; grounding should make source authority and context explicit; the present case is different and calls for the team to use structure-aware boundaries and suitable overlap between adjacent chunks.
B:Source metadata is application evidence; unconstrained text generation can fabricate bibliographic details in its own workflow, but that is a different correction from the present requirement to use structure-aware boundaries and suitable overlap between adjacent chunks. The reported issue would remain.
D:That change addresses another failure mode: a semantic match can still be stale or out of scope; retrieval must honor business applicability; it does not address this case, for which the relevant intervention is to use structure-aware boundaries and suitable overlap between adjacent chunks.
Question 159
New documents enter a vector index but have no populated vector field. What indexing step is missing?
- Pass authoritative current sources with clear attribution and precedence instructions
- Filter to the currently effective, access-authorized policy version
- Run embedding generation for the new content before expecting vector-neighbor retrieval
- Construct citations from actual retrieved source metadata rather than free-form model guesses
Explanation
Correct Answer
C:Text fields alone do not create embeddings unless supported vectorization is configured. In the private-document answer assistant, confirm the fix by retesting the affected behavior after running embedding generation for the new content before expecting vector-neighbor retrieval.
Incorrect Answers
A:A pile of unqualified passages can encourage unsupported synthesis; grounding should make source authority and context explicit in its own workflow, but that is a different correction from the present requirement to run embedding generation for the new content before expecting vector-neighbor retrieval. The reported issue would remain.
B:Although useful elsewhere in the private-document answer assistant, a semantic match can still be stale or out of scope; retrieval must honor business applicability; the present case is different and calls for the team to run embedding generation for the new content before expecting vector-neighbor retrieval.
D:That change addresses another failure mode: source metadata is application evidence; unconstrained text generation can fabricate bibliographic details; it does not address this case, for which the relevant intervention is to run embedding generation for the new content before expecting vector-neighbor retrieval.
Question 160
The embedding model accepts only bounded input length, while some source chunks exceed the limit. What is needed?
- Require a supported-answer fallback when the grounding set is empty or insufficient
- Reduce chunk size before embedding and preserve document-to-chunk identifiers
- Interpret similarity as retrieval relevance rather than a guarantee that a passage is correct
- Record source version or retrieval timestamp alongside the cited document identifier
Explanation
Correct Answer
B:Embedding input limits are model constraints; truncating oversized passages silently may discard important evidence. In the private-document answer assistant, confirm the fix by retesting the affected behavior after reducing chunk size before embedding and preserve document-to-chunk identifiers.
Incorrect Answers
A:Although useful elsewhere in the private-document answer assistant, rAG should not manufacture facts in the absence of evidence; an explicit abstention path is safer; the present case is different and calls for the team to reduce chunk size before embedding and preserve document-to-chunk identifiers.
C:That change addresses another failure mode: embedding proximity estimates semantic relatedness, not the authenticity or correctness of a source claim; it does not address this case, for which the relevant intervention is to reduce chunk size before embedding and preserve document-to-chunk identifiers.
D:Versioned provenance helps determine which source text supported an earlier answer in its own workflow, but that is a different correction from the present requirement to reduce chunk size before embedding and preserve document-to-chunk identifiers. The reported issue would remain.