{"id":18159,"date":"2026-09-22T05:41:59","date_gmt":"2026-09-22T05:41:59","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=18159"},"modified":"2026-09-22T05:41:59","modified_gmt":"2026-09-22T05:41:59","slug":"databricks-certified-generative-ai-engineer-associate-practice-test-questions-and-exam-dumps-part19-q361-380","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/databricks-certified-generative-ai-engineer-associate-practice-test-questions-and-exam-dumps-part19-q361-380\/","title":{"rendered":"Databricks Certified Generative AI Engineer Associate Practice Test Questions and Exam Dumps Part19 Q361-380"},"content":{"rendered":"

View Full <\/b>Databricks Certified Generative AI Engineer Associate Exam Dumps<\/b><\/a> and Practice Test Dumps<\/b><\/p>\n

 <\/p>\n

Question 361. Which component is responsible for generating the final natural-language response in a typical RAG architecture?<\/b><\/p>\n

    \n
  1. <\/b> Vector index<\/span>
    \n<\/span>2.<\/b> Metadata filter<\/span>
    \n<\/span>3.<\/b> Language model<\/span>
    \n<\/span>4.<\/b> Document loader<\/span><\/li>\n<\/ol>\n

    Correct Answer: 3. Language model<\/b><\/p>\n

    Explanation:<\/b><\/p>\n

    The language model is responsible for generating the final response after the application has assembled the relevant context. In a RAG workflow, the retriever first identifies useful information from an external knowledge source. That information is then placed into the model’s prompt along with the user’s request and appropriate instructions. The language model uses this context to produce a natural-language response. Although the model performs generation, its answer quality depends heavily on the quality of retrieved evidence, prompt construction, and application controls. A complete RAG system therefore combines retrieval and generation rather than expecting the language model alone to provide all required factual information.<\/span><\/p>\n

    Question 362. What is a primary benefit of using metadata alongside document chunks?<\/b><\/p>\n

      \n
    1. <\/b> It allows filtering and additional context during retrieval<\/span>
      \n<\/span>2.<\/b> It eliminates the need for embeddings<\/span>
      \n<\/span>3.<\/b> It automatically increases model intelligence<\/span>
      \n<\/span>4.<\/b> It guarantees hallucination-free responses<\/span><\/li>\n<\/ol>\n

      Correct Answer: 1. It allows filtering and additional context during retrieval<\/b><\/p>\n

      Explanation:<\/b><\/p>\n

      Metadata provides structured information associated with each document or chunk. Examples include document type, source, department, publication date, effective date, version, language, access classification, or product category. Retrieval systems can use this information to filter candidates before or during semantic search. Metadata can also help the application determine which source should take precedence when multiple versions exist. Keeping useful metadata with chunks makes retrieval more precise and enables additional application logic that semantic similarity alone cannot provide. Metadata quality is important because incorrect or missing attributes can cause relevant information to be excluded or inappropriate documents to be retrieved.<\/span><\/p>\n

      Question 363. Which approach is most appropriate for a RAG application that must answer questions using confidential documents?<\/b><\/p>\n

        \n
      1. <\/b> Provide all documents to every user<\/span>
        \n<\/span>2.<\/b> Apply authorization controls before confidential content reaches the model<\/span>
        \n<\/span>3.<\/b> Increase the model temperature<\/span>
        \n<\/span>4.<\/b> Remove document ownership metadata<\/span><\/li>\n<\/ol>\n

        Correct Answer: 2. Apply authorization controls before confidential content reaches the model<\/b><\/p>\n

        Explanation:<\/b><\/p>\n

        Confidential information should be protected through explicit access-control mechanisms before it is included in the model’s context. Retrieval can use authorization metadata to ensure that only documents the requesting user is permitted to access become candidates for the response. Relying only on instructions telling the model not to reveal sensitive information is insufficient because the restricted information has already entered the model context. Authorization should therefore be enforced at the application or data layer. Security testing should also verify that filtering cannot be bypassed through alternative queries. This approach helps protect confidential information while still allowing authorized users to benefit from enterprise knowledge.<\/span><\/p>\n

        Question 364. What can happen if a RAG application sends an entire large document to the model instead of retrieving relevant sections?<\/b><\/p>\n

          \n
        1. <\/b> It always improves answer quality<\/span>
          \n<\/span>2.<\/b> It can increase token usage and introduce irrelevant information<\/span>
          \n<\/span>3.<\/b> It removes the need for context windows<\/span>
          \n<\/span>4.<\/b> It guarantees better retrieval recall<\/span><\/li>\n<\/ol>\n

          Correct Answer: 2. It can increase token usage and introduce irrelevant information<\/b><\/p>\n

          Explanation:<\/b><\/p>\n

          Passing an entire large document to the language model may provide more information than is necessary for the user’s question. This increases token consumption and can increase inference latency and cost. It can also introduce unrelated material that distracts the model from the evidence relevant to the actual request. RAG systems generally aim to retrieve focused sections or chunks so that the model receives useful context without unnecessary content. However, retrieval must still preserve enough surrounding information to support the answer. Chunking, metadata filtering, reranking, and appropriate top-k selection can help construct a focused context window.<\/span><\/p>\n

          Question 365. Why is a representative evaluation dataset important for generative AI applications?<\/b><\/p>\n

            \n
          1. <\/b> It makes all models behave identically<\/span>
            \n<\/span>2.<\/b> It removes the need for monitoring<\/span>
            \n<\/span>3.<\/b> It reflects the types of queries and scenarios expected in actual use<\/span>
            \n<\/span>4.<\/b> It automatically fixes retrieval errors<\/span><\/li>\n<\/ol>\n

            Correct Answer: 3. It reflects the types of queries and scenarios expected in actual use<\/b><\/p>\n

            Explanation:<\/b><\/p>\n

            A representative evaluation dataset provides realistic test cases that reflect how users are expected to interact with the application. It can include common questions, difficult queries, ambiguous requests, domain terminology, edge cases, and situations where insufficient evidence should be detected. Using representative data makes evaluation results more meaningful because improvements measured on unrelated examples may not translate into better production behavior. The dataset should remain sufficiently stable to support comparisons between versions. It can be combined with automated metrics and human review to assess retrieval quality, groundedness, relevance, correctness, and other characteristics that matter for the application’s intended use.<\/span><\/p>\n

            Question 366. What is the purpose of a content hash during document ingestion?<\/b><\/p>\n

              \n
            1. <\/b> To detect whether document content has changed<\/span>
              \n<\/span>2.<\/b> To increase the model’s context window<\/span>
              \n<\/span>3.<\/b> To replace authorization<\/span>
              \n<\/span>4.<\/b> To generate the final answer<\/span><\/li>\n<\/ol>\n

              Correct Answer: 1. To detect whether document content has changed<\/b><\/p>\n

              Explanation:<\/b><\/p>\n

              A content hash can provide a compact representation of document content that allows an ingestion pipeline to identify changes efficiently. If the hash of a newly observed document matches the previously stored hash, the system can determine that the content has not changed and may avoid unnecessary reprocessing. If the hash differs, the document can be re-chunked, re-embedded, and updated in the retrieval index as appropriate. This is particularly useful for large knowledge bases where only a small percentage of documents change between ingestion cycles. Content hashes can therefore support efficient incremental ingestion while reducing unnecessary computation.<\/span><\/p>\n

              Question 367. Which problem can occur when chunks are too small?<\/b><\/p>\n

                \n
              1. <\/b> Important contextual relationships may be separated<\/span>
                \n<\/span>2.<\/b> Retrieval becomes impossible<\/span>
                \n<\/span>3.<\/b> The model automatically becomes more accurate<\/span>
                \n<\/span>4.<\/b> Metadata can no longer be stored<\/span><\/li>\n<\/ol>\n

                Correct Answer: 1. Important contextual relationships may be separated<\/b><\/p>\n

                Explanation:<\/b><\/p>\n

                Very small chunks can improve precision in some situations, but they may also remove important context needed to understand the retrieved information. A sentence containing a key statement may depend on definitions, qualifications, headings, or previous sentences that were placed into another chunk. If only the small fragment is retrieved, the language model may not have enough evidence to interpret it correctly. Chunk size should therefore balance focused retrieval with contextual completeness. Document structure, semantic boundaries, and overlap can help preserve useful relationships. The appropriate strategy should be tested against representative queries because different document types have different contextual requirements.<\/span><\/p>\n

                Question 368. What is the main purpose of response relevance evaluation?<\/b><\/p>\n

                  \n
                1. <\/b> To determine whether the generated answer appropriately addresses the user’s request<\/span>
                  \n<\/span>2.<\/b> To measure vector index storage capacity<\/span>
                  \n<\/span>3.<\/b> To determine document file size<\/span>
                  \n<\/span>4.<\/b> To verify network bandwidth<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 1. To determine whether the generated answer appropriately addresses the user’s request<\/b><\/p>\n

                  Explanation:<\/b><\/p>\n

                  Response relevance evaluates whether the generated answer actually addresses the user’s question or task. A response can be grammatically correct and grounded in retrieved information while still failing to answer what the user asked. For example, the model may provide related background information but omit the requested detail. Relevance evaluation therefore complements metrics such as groundedness, which focuses on support from evidence. Evaluating relevance can involve automated methods, reference-based comparisons, or human review depending on the application. A representative evaluation dataset helps determine whether changes to prompts, retrieval, or model configurations improve the usefulness of responses.<\/span><\/p>\n

                  Question 369. Which strategy can help reduce hallucinations when the retrieval system cannot find sufficient evidence?<\/b><\/p>\n

                    \n
                  1. <\/b> Encourage the model to make an educated guess<\/span>
                    \n<\/span>2.<\/b> Instruct the model to acknowledge insufficient evidence<\/span>
                    \n<\/span>3.<\/b> Increase temperature<\/span>
                    \n<\/span>4.<\/b> Retrieve random documents<\/span><\/li>\n<\/ol>\n

                    Correct Answer: 2. Instruct the model to acknowledge insufficient evidence<\/b><\/p>\n

                    Explanation:<\/b><\/p>\n

                    When a retrieval system cannot provide adequate evidence, encouraging the model to guess can lead to unsupported claims. A safer strategy is to explicitly instruct the model to acknowledge when the available context does not support an answer. The application can also use similarity thresholds to detect weak retrieval results and decide whether to answer, ask for clarification, or indicate that sufficient information was not found. This approach improves transparency and reduces the likelihood that the system will present unsupported information as fact. The behavior should be evaluated using test cases specifically designed to represent missing, ambiguous, or insufficient knowledge.<\/span><\/p>\n

                    Question 370. What is a benefit of using hybrid retrieval in a knowledge base containing both general descriptions and exact identifiers?<\/b><\/p>\n

                      \n
                    1. <\/b> It combines semantic and lexical retrieval strengths<\/span>
                      \n<\/span>2.<\/b> It eliminates all retrieval errors<\/span>
                      \n<\/span>3.<\/b> It removes the need for document metadata<\/span>
                      \n<\/span>4.<\/b> It prevents documents from being updated<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 1. It combines semantic and lexical retrieval strengths<\/b><\/p>\n

                      Explanation:<\/b><\/p>\n

                      Hybrid retrieval combines semantic similarity with lexical or keyword matching. This can be particularly useful when a knowledge base contains both descriptive content and exact identifiers such as product codes, ticket numbers, error messages, command names, or version strings. Semantic search can capture related concepts even when wording differs, while lexical search can identify exact terms that are important to the query. Combining the approaches can improve retrieval coverage across different query types. The results may then be merged or reranked before being provided to the language model. Evaluation should determine whether hybrid retrieval improves the specific application’s precision and recall.<\/span><\/p>\n

                      Question 371. Why should an application track the version of the embedding model used for a vector index?<\/b><\/p>\n

                        \n
                      1. <\/b> To identify which embedding configuration produced the indexed vectors<\/span>
                        \n<\/span>2.<\/b> To increase generation temperature<\/span>
                        \n<\/span>3.<\/b> To eliminate metadata filtering<\/span>
                        \n<\/span>4.<\/b> To replace the language model<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 1. To identify which embedding configuration produced the indexed vectors<\/b><\/p>\n

                        Explanation:<\/b><\/p>\n

                        The embedding model influences the numerical representation of indexed documents and queries. If the embedding model changes, the characteristics of the resulting vectors may also change, and previously generated document embeddings may no longer be directly compatible with the new query embeddings. Tracking the embedding model version helps teams understand how an index was created and reproduce or troubleshoot retrieval behavior. If a new embedding model is adopted, the knowledge base may need to be re-embedded so that document and query representations are generated consistently. Version tracking therefore supports reproducibility, controlled experimentation, and reliable retrieval operations.<\/span><\/p>\n

                        Question 372. Which practice can help ensure that a RAG application uses authoritative information when sources conflict?<\/b><\/p>\n

                          \n
                        1. <\/b> Randomly choose one source<\/span>
                          \n<\/span>2.<\/b> Use source-authority metadata and business rules<\/span>
                          \n<\/span>3.<\/b> Increase model temperature<\/span>
                          \n<\/span>4.<\/b> Remove publication dates<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 2. Use source-authority metadata and business rules<\/b><\/p>\n

                          Explanation:<\/b><\/p>\n

                          When multiple sources contain conflicting information, the application needs explicit rules for deciding which evidence should be preferred. Metadata can identify authoritative sources, document owners, publication status, effective dates, or approved versions. Retrieval can then filter or rank documents according to these business rules. This is safer than expecting the language model to determine authority solely from document content. For example, an approved internal policy may take precedence over an archived draft. If the system cannot confidently resolve the conflict, the application should communicate the uncertainty rather than presenting an unsupported conclusion. Source authority should be maintained as part of the knowledge management process.<\/span><\/p>\n

                          Question 373. What does retrieval precision measure?<\/b><\/p>\n

                            \n
                          1. <\/b> The proportion of retrieved results that are relevant<\/span>
                            \n<\/span>2.<\/b> The number of documents stored in the index<\/span>
                            \n<\/span>3.<\/b> The number of model parameters<\/span>
                            \n<\/span>4.<\/b> The response generation time<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 1. The proportion of retrieved results that are relevant<\/b><\/p>\n

                            Explanation:<\/b><\/p>\n

                            Retrieval precision measures how many of the documents returned by a retrieval system are actually relevant to the query. High precision means that the retrieval results contain relatively little irrelevant information. This is valuable in RAG because irrelevant documents can increase context size and distract the language model from the strongest evidence. Precision should be considered alongside recall because a system can achieve high precision by returning very few documents while missing other relevant information. Techniques such as reranking, metadata filtering, query rewriting, and similarity thresholds can help improve precision while maintaining sufficient recall.<\/span><\/p>\n

                            Question 374. Why can user feedback be useful in improving a production RAG application?<\/b><\/p>\n

                              \n
                            1. <\/b> It provides real-world signals about response usefulness and failure cases<\/span>
                              \n<\/span>2.<\/b> It automatically changes model weights<\/span>
                              \n<\/span>3.<\/b> It eliminates evaluation datasets<\/span>
                              \n<\/span>4.<\/b> It guarantees accurate retrieval<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 1. It provides real-world signals about response usefulness and failure cases<\/b><\/p>\n

                              Explanation:<\/b><\/p>\n

                              User feedback can reveal issues that may not appear in controlled evaluation datasets. Users may identify answers that are technically relevant but incomplete, unclear, outdated, poorly grounded, or missing important information. Feedback can be categorized and incorporated into future evaluation datasets or improvement workflows. For example, repeated complaints about a particular type of query may indicate a retrieval problem, missing source data, or a prompt issue. Feedback should be interpreted carefully because individual responses may be subjective or incomplete. Combining user feedback with automated metrics, traces, and structured evaluation provides a stronger basis for deciding which parts of the system need improvement.<\/span><\/p>\n

                              Question 375. What is a major advantage of using structured metadata filters before semantic retrieval?<\/b><\/p>\n

                                \n
                              1. <\/b> They can narrow the search space according to known constraints<\/span>
                                \n<\/span>2.<\/b> They increase the model’s parameter count<\/span>
                                \n<\/span>3.<\/b> They eliminate the need for embeddings in every application<\/span>
                                \n<\/span>4.<\/b> They guarantee perfect responses<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 1. They can narrow the search space according to known constraints<\/b><\/p>\n

                                Explanation:<\/b><\/p>\n

                                Metadata filters can apply known constraints such as document type, department, date, language, product, region, or access level before or during retrieval. This can significantly reduce irrelevant candidates and improve retrieval precision. For example, a query about a current product manual can be restricted to the relevant product and active documentation before semantic ranking is applied. Structured filtering is complementary to semantic retrieval rather than a complete replacement for it. The effectiveness of metadata filters depends on the quality and consistency of the metadata stored during ingestion. Poor metadata can cause relevant documents to be incorrectly excluded.<\/span><\/p>\n

                                Question 376. Which action should occur before passing retrieved documents to the language model in a secure enterprise RAG application?<\/b><\/p>\n

                                  \n
                                1. <\/b> Apply appropriate authorization and relevance controls<\/span>
                                  \n<\/span>2.<\/b> Increase temperature<\/span>
                                  \n<\/span>3.<\/b> Remove all source identifiers<\/span>
                                  \n<\/span>4.<\/b> Ignore document metadata<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 1. Apply appropriate authorization and relevance controls<\/b><\/p>\n

                                  Explanation:<\/b><\/p>\n

                                  Before retrieved content is included in a model prompt, the application should ensure that the documents are both relevant to the query and authorized for the requesting user. Authorization controls prevent confidential or restricted information from entering the model context, while relevance controls reduce unnecessary or unrelated information. Depending on the architecture, these controls can include metadata filters, access-control rules, similarity thresholds, reranking, and source-authority checks. Applying these controls before generation is important because the model should not be expected to enforce permissions after sensitive information has already been supplied. Secure context construction is therefore an essential part of enterprise RAG design.<\/span><\/p>\n

                                  Question 377. What is one reason to use a baseline before experimenting with a new RAG configuration?<\/b><\/p>\n

                                    \n
                                  1. <\/b> It provides a reference for measuring whether the change improves performance<\/span>
                                    \n<\/span>2.<\/b> It prevents any future changes<\/span>
                                    \n<\/span>3.<\/b> It makes retrieval unnecessary<\/span>
                                    \n<\/span>4.<\/b> It guarantees production success<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 1. It provides a reference for measuring whether the change improves performance<\/b><\/p>\n

                                    Explanation:<\/b><\/p>\n

                                    A baseline represents a known configuration and its measured performance before a change is introduced. It can include retrieval precision and recall, groundedness, response relevance, latency, cost, or other application-specific metrics. When a new configuration is tested using the same evaluation conditions, results can be compared with the baseline to identify improvements or regressions. This is more reliable than judging changes solely from individual examples or subjective impressions. Baseline configurations should be documented and versioned so they can be reproduced when needed. A strong baseline provides an objective starting point for iterative optimization of prompts, models, retrieval, and data.<\/span><\/p>\n

                                    Question 378. Which practice helps reduce the chance of serving stale information after source documents are updated?<\/b><\/p>\n

                                      \n
                                    1. <\/b> Maintain synchronization between source data and the retrieval index<\/span>
                                      \n<\/span>2.<\/b> Ignore document updates<\/span>
                                      \n<\/span>3.<\/b> Increase model temperature<\/span>
                                      \n<\/span>4.<\/b> Keep obsolete embeddings permanently<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 1. Maintain synchronization between source data and the retrieval index<\/b><\/p>\n

                                      Explanation:<\/b><\/p>\n

                                      A RAG application can only retrieve current information if its index reflects the current state of the authoritative source data. When documents are added, modified, or removed, the ingestion process should detect these changes and update the corresponding chunks, embeddings, metadata, and index entries. Incremental ingestion can make this process efficient by processing only changed content. Version identifiers, timestamps, and content hashes can help detect updates. Monitoring ingestion freshness can also provide visibility into synchronization failures. Without an effective update process, the model may continue receiving obsolete evidence even though the original source system already contains newer information.<\/span><\/p>\n

                                      Question 379. What is the purpose of tracing individual RAG requests?<\/b><\/p>\n

                                        \n
                                      1. <\/b> To understand the sequence of operations that produced a response<\/span>
                                        \n<\/span>2.<\/b> To increase model context automatically<\/span>
                                        \n<\/span>3.<\/b> To replace all evaluation metrics<\/span>
                                        \n<\/span>4.<\/b> To prevent documents from changing<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 1. To understand the sequence of operations that produced a response<\/b><\/p>\n

                                        Explanation:<\/b><\/p>\n

                                        Tracing provides a detailed view of the operations involved in processing a request. A trace can connect the original user query with query rewriting, embedding generation, retrieval, filtering, reranking, prompt construction, model generation, and output validation. This makes it easier to identify where a failure occurred. For example, a poor answer might result from an incorrect query rewrite, missing retrieval evidence, inappropriate filtering, or generation behavior. Tracing can also help analyze latency across individual stages. Because traces may contain sensitive information, access controls, retention policies, and data minimization should be considered when designing production observability.<\/span><\/p>\n

                                        Question 380. Which sequence best represents a controlled improvement cycle for a RAG application?<\/b><\/p>\n

                                          \n
                                        1. <\/b> Change several components, deploy immediately, and evaluate later<\/span>
                                          \n<\/span>2.<\/b> Collect observations, form a hypothesis, make a controlled change, evaluate, and monitor<\/span>
                                          \n<\/span>3.<\/b> Remove the baseline and rely on user opinions<\/span>
                                          \n<\/span>4.<\/b> Change the model temperature whenever quality decreases<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 2. Collect observations, form a hypothesis, make a controlled change, evaluate, and monitor<\/b><\/p>\n

                                          Explanation:<\/b><\/p>\n

                                          A controlled improvement cycle begins by identifying evidence of a problem through evaluation, monitoring, traces, or user feedback. The team can then form a hypothesis about the likely cause and make a targeted change to one or more well-defined components. The modified configuration should be evaluated against a stable and representative dataset using relevant metrics. If the change performs well, it can be deployed with production monitoring and retained as a versioned configuration. If performance regresses, the previous validated configuration can be restored. This disciplined cycle makes improvements measurable, reproducible, and easier to troubleshoot than untracked changes made directly in production.<\/span><\/p>\n

                                           <\/p>\n","protected":false},"excerpt":{"rendered":"

                                          View Full Databricks Certified Generative AI Engineer Associate Exam Dumps and Practice Test Dumps   Question 361. Which component is responsible for generating the final natural-language response in a typical RAG architecture? Vector index 2. Metadata filter 3. Language model 4. Document loader Correct Answer: 3. Language model Explanation: The language model is responsible for […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[1648,1647],"tags":[],"_links":{"self":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/18159"}],"collection":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/comments?post=18159"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/18159\/revisions"}],"predecessor-version":[{"id":18160,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/18159\/revisions\/18160"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=18159"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=18159"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=18159"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}