{"id":18157,"date":"2026-09-22T05:41:32","date_gmt":"2026-09-22T05:41:32","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=18157"},"modified":"2026-09-22T05:41:32","modified_gmt":"2026-09-22T05:41:32","slug":"databricks-certified-generative-ai-engineer-associate-practice-test-questions-and-exam-dumps-part18-q341-360","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/databricks-certified-generative-ai-engineer-associate-practice-test-questions-and-exam-dumps-part18-q341-360\/","title":{"rendered":"Databricks Certified Generative AI Engineer Associate Practice Test Questions and Exam Dumps Part18 Q341-360"},"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 341. What is the primary purpose of a system prompt in a generative AI application?<\/b><\/p>\n

    \n
  1. <\/b> To store document embeddings<\/span>
    \n<\/span>2.<\/b> To define high-level instructions and behavior for the model<\/span>
    \n<\/span>3.<\/b> To replace the retrieval index<\/span>
    \n<\/span>4.<\/b> To increase network bandwidth<\/span><\/li>\n<\/ol>\n

    Correct Answer: 2. To define high-level instructions and behavior for the model<\/b><\/p>\n

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

    A system prompt provides high-level instructions that establish how a language model should behave when responding to requests. In a RAG application, these instructions can specify the model’s role, response style, grounding requirements, handling of insufficient evidence, and formatting expectations. A system prompt can also instruct the model to use retrieved context rather than relying on unsupported assumptions. However, prompt instructions should not be treated as a substitute for application-level security or authorization controls. System prompts are one component of a larger architecture that can include retrieval, filtering, validation, monitoring, and other safeguards.<\/span><\/p>\n

    Question 342. Which component converts a user’s text query into a representation that can be compared with document embeddings?<\/b><\/p>\n

      \n
    1. <\/b> Output validator<\/span>
      \n<\/span>2.<\/b> Reranker<\/span>
      \n<\/span>3.<\/b> Embedding model<\/span>
      \n<\/span>4.<\/b> Logging service<\/span><\/li>\n<\/ol>\n

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

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

      An embedding model transforms text into a numerical vector representation that captures semantic characteristics of the input. In a RAG workflow, the user’s query is converted into an embedding and compared with embeddings representing indexed document chunks. Similarity calculations can then identify documents that are semantically related to the query. The embedding model therefore plays a critical role in semantic retrieval but does not normally generate the final natural-language response. Document and query embeddings should generally be compatible, meaning the same or appropriately compatible embedding approach should be used when indexing and querying. Retrieval quality should be evaluated using representative application data.<\/span><\/p>\n

      Question 343. What is a major risk of using outdated documents as retrieval sources?<\/b><\/p>\n

        \n
      1. <\/b> The model may generate answers based on superseded information<\/span>
        \n<\/span>2.<\/b> The model automatically becomes smaller<\/span>
        \n<\/span>3.<\/b> The vector index becomes unnecessary<\/span>
        \n<\/span>4.<\/b> Tokenization stops working<\/span><\/li>\n<\/ol>\n

        Correct Answer: 1. The model may generate answers based on superseded information<\/b><\/p>\n

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

        Outdated documents can cause a RAG application to provide information that is no longer valid even when the retrieval process technically succeeds. For example, an old company policy or product specification may be retrieved because it is semantically relevant to the query, but a newer approved version may have replaced it. This is why document lifecycle management, version metadata, effective dates, and source authority are important. Ingestion pipelines should identify obsolete content and either remove it, mark it appropriately, or prevent it from being selected for current queries. Retrieval quality therefore depends not only on semantic similarity but also on the freshness and authority of the underlying knowledge.<\/span><\/p>\n

        Question 344. Which technique can improve retrieval when exact product codes or error identifiers are important?<\/b><\/p>\n

          \n
        1. <\/b> Increase model temperature<\/span>
          \n<\/span>2.<\/b> Use lexical or hybrid search<\/span>
          \n<\/span>3.<\/b> Remove document metadata<\/span>
          \n<\/span>4.<\/b> Reduce all chunks to single words<\/span><\/li>\n<\/ol>\n

          Correct Answer: 2. Use lexical or hybrid search<\/b><\/p>\n

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

          Exact product codes, error identifiers, command names, and version numbers can be difficult for semantic retrieval alone because users may need an exact textual match. Lexical search is designed to identify matching terms, while hybrid retrieval combines lexical matching with semantic similarity. This combination can provide broader retrieval coverage for technical queries. For example, a user searching for a specific error code can benefit from exact keyword matching, while a user asking what that error means can benefit from semantic retrieval. The appropriate approach should be validated using representative queries because different document collections may require different retrieval strategies.<\/span><\/p>\n

          Question 345. What is the purpose of a metadata filter that restricts documents to a particular department?<\/b><\/p>\n

            \n
          1. <\/b> To increase model temperature<\/span>
            \n<\/span>2.<\/b> To reduce embedding dimensions<\/span>
            \n<\/span>3.<\/b> To limit retrieval to documents matching the specified department<\/span>
            \n<\/span>4.<\/b> To disable vector search<\/span><\/li>\n<\/ol>\n

            Correct Answer: 3. To limit retrieval to documents matching the specified department<\/b><\/p>\n

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

            Metadata filters allow retrieval systems to apply structured conditions before or during document selection. A department field can restrict retrieval to information associated with a particular business unit, reducing irrelevant results from other areas. Metadata filtering can also support security, document type restrictions, geographic scope, effective dates, and other application-specific requirements. When combined with semantic similarity, filtering can produce a more focused retrieval set than semantic search alone. The metadata itself must be accurate and consistently maintained during ingestion. Incorrect metadata can cause useful information to be excluded or inappropriate information to be included.<\/span><\/p>\n

            Question 346. Why is document structure useful when designing a chunking strategy?<\/b><\/p>\n

              \n
            1. <\/b> It can help preserve logical relationships between related content<\/span>
              \n<\/span>2.<\/b> It automatically increases model parameters<\/span>
              \n<\/span>3.<\/b> It removes the need for embeddings<\/span>
              \n<\/span>4.<\/b> It guarantees perfect retrieval<\/span><\/li>\n<\/ol>\n

              Correct Answer: 1. It can help preserve logical relationships between related content<\/b><\/p>\n

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

              Document structure can provide valuable signals for determining where chunks should begin and end. Headings, sections, paragraphs, tables, lists, and other structural elements often represent meaningful boundaries. If a chunking strategy ignores these boundaries, related information may be split in ways that reduce its usefulness during retrieval. For example, separating a heading from the explanation that follows it can make the retrieved content harder to interpret. Structure-aware chunking can preserve semantic relationships while keeping chunks sufficiently focused. The resulting strategy should still be evaluated because different document types may require different approaches, and no single chunking method is optimal for every knowledge base.<\/span><\/p>\n

              Question 347. What is the purpose of query rewriting in a multi-turn RAG application?<\/b><\/p>\n

                \n
              1. <\/b> To convert the current request into a self-contained retrieval query using conversation context<\/span>
                \n<\/span>2.<\/b> To replace the language model<\/span>
                \n<\/span>3.<\/b> To delete all conversation history<\/span>
                \n<\/span>4.<\/b> To increase the vector index size<\/span><\/li>\n<\/ol>\n

                Correct Answer: 1. To convert the current request into a self-contained retrieval query using conversation context<\/b><\/p>\n

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

                In multi-turn conversations, users often ask follow-up questions using pronouns or references such as \u201cthat product,\u201d \u201cthe previous policy,\u201d or \u201cwhat about the second option.\u201d A retrieval system may struggle with these references if it receives only the latest message. Query rewriting can use relevant conversation history to create a self-contained query that expresses the user’s current intent more clearly. This rewritten query can then be sent to the retrieval system. The technique can improve retrieval accuracy without requiring the retrieval index to understand the entire conversation. The rewriting process should preserve the user’s intent and avoid introducing unsupported assumptions.<\/span><\/p>\n

                Question 348. Which practice helps reduce the impact of prompt changes on production reliability?<\/b><\/p>\n

                  \n
                1. <\/b> Change prompts without recording versions<\/span>
                  \n<\/span>2.<\/b> Version, evaluate, and monitor prompt changes<\/span>
                  \n<\/span>3.<\/b> Remove evaluation datasets<\/span>
                  \n<\/span>4.<\/b> Disable production monitoring<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 2. Version, evaluate, and monitor prompt changes<\/b><\/p>\n

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

                  Prompt changes can alter model behavior even when the underlying model and retrieval system remain unchanged. A revised prompt may improve groundedness or formatting while unexpectedly reducing response relevance or changing how insufficient evidence is handled. Versioning prompts provides a record of what changed, while evaluation against a stable dataset provides evidence about the effect of that change. Production monitoring can then identify behavior that was not captured during testing. This combination supports controlled iteration and makes rollback possible when necessary. Treating prompts as versioned application artifacts is therefore useful for maintaining reproducibility and production reliability.<\/span><\/p>\n

                  Question 349. Which metric is most directly concerned with whether a generated answer is supported by retrieved evidence?<\/b><\/p>\n

                    \n
                  1. <\/b> Throughput<\/span>
                    \n<\/span>2.<\/b> Retrieval recall<\/span>
                    \n<\/span>3.<\/b> Groundedness<\/span>
                    \n<\/span>4.<\/b> Network latency<\/span><\/li>\n<\/ol>\n

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

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

                    Groundedness evaluates whether the claims made in a generated response are supported by the information supplied as evidence. In a RAG application, retrieved documents are intended to provide the factual basis for the model’s response. A response can be relevant to a user’s question while still containing unsupported claims, so relevance and groundedness measure different aspects of quality. Groundedness evaluation can help identify cases where the model introduces information that was not present in the retrieved context. It is useful alongside retrieval metrics and human evaluation because reliable responses depend on both obtaining appropriate evidence and using that evidence correctly.<\/span><\/p>\n

                    Question 350. What is one reason to use a reranker after initial vector retrieval?<\/b><\/p>\n

                      \n
                    1. <\/b> To perform a more detailed relevance assessment on candidate documents<\/span>
                      \n<\/span>2.<\/b> To remove the need for document embeddings<\/span>
                      \n<\/span>3.<\/b> To increase model context indefinitely<\/span>
                      \n<\/span>4.<\/b> To replace access controls<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 1. To perform a more detailed relevance assessment on candidate documents<\/b><\/p>\n

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

                      Initial vector retrieval is often designed to quickly identify a broad set of potentially relevant candidates. Some candidates may be only partially related to the user’s query, especially when the knowledge base contains many semantically similar documents. A reranker can evaluate the query and retrieved candidates more carefully and reorder them according to relevance. The application can then select the strongest results for final context construction. This approach can improve the quality of evidence without requiring the initial retrieval stage to perform expensive detailed comparisons across the entire knowledge base. Reranking should still be evaluated because its benefits depend on the quality of the candidate set and the application domain.<\/span><\/p>\n

                      Question 351. Why should retrieved context be limited to information relevant to the user’s request?<\/b><\/p>\n

                        \n
                      1. <\/b> Irrelevant context can increase noise, token usage, and model confusion<\/span>
                        \n<\/span>2.<\/b> The model cannot process any context<\/span>
                        \n<\/span>3.<\/b> Retrieval becomes unnecessary<\/span>
                        \n<\/span>4.<\/b> It guarantees shorter documents<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 1. Irrelevant context can increase noise, token usage, and model confusion<\/b><\/p>\n

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

                        Providing too much irrelevant context can make a RAG response less reliable. Additional documents consume context-window capacity and can increase inference cost and latency. More importantly, irrelevant or conflicting information may distract the model from the evidence that actually answers the user’s question. Retrieval systems should therefore aim to provide a focused set of high-quality evidence rather than simply maximizing the amount of context. Techniques such as metadata filtering, similarity thresholds, reranking, and appropriate top-k settings can help achieve this balance. The ideal amount of context depends on the task and should be determined through evaluation rather than assumed.<\/span><\/p>\n

                        Question 352. Which feature is particularly useful for handling structured responses from a language model?<\/b><\/p>\n

                          \n
                        1. <\/b> Output schema validation<\/span>
                          \n<\/span>2.<\/b> Increasing chunk overlap<\/span>
                          \n<\/span>3.<\/b> Removing metadata<\/span>
                          \n<\/span>4.<\/b> Increasing retrieval top-k indefinitely<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 1. Output schema validation<\/b><\/p>\n

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

                          Output schema validation checks whether the model’s response follows a predefined structural format required by the application. For example, an application may expect a JSON object containing a category, explanation, confidence field, and source references. Even when a prompt requests a specific structure, models can occasionally omit fields or produce invalid data types. Validation provides a programmatic checkpoint before the output reaches downstream systems. If the output fails validation, the application can reject it, request a corrected response, or use another handling strategy. Schema validation improves integration reliability, although it does not by itself guarantee factual accuracy or groundedness.<\/span><\/p>\n

                          Question 353. What is an important consideration when using conversation history as retrieval context?<\/b><\/p>\n

                            \n
                          1. <\/b> Irrelevant or excessive history can consume context and introduce noise<\/span>
                            \n<\/span>2.<\/b> All history must always be included<\/span>
                            \n<\/span>3.<\/b> Conversation history replaces document retrieval<\/span>
                            \n<\/span>4.<\/b> History automatically guarantees correct answers<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 1. Irrelevant or excessive history can consume context and introduce noise<\/b><\/p>\n

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

                            Conversation history can provide useful context for understanding follow-up questions, but sending the entire history to every request may become inefficient. Long conversations consume context-window capacity and can include outdated or irrelevant information. This can make retrieval and generation more difficult. Applications can manage history by selecting relevant messages, summarizing older interactions, or rewriting the current query using only the necessary context. The strategy should preserve information that affects the user’s current intent while avoiding unnecessary conversational details. Effective history management can improve both retrieval quality and operational efficiency in multi-turn generative AI applications.<\/span><\/p>\n

                            Question 354. Which approach is useful when documents contain multiple versions of the same policy?<\/b><\/p>\n

                              \n
                            1. <\/b> Ignore all version information<\/span>
                              \n<\/span>2.<\/b> Use version and effective-date metadata<\/span>
                              \n<\/span>3.<\/b> Increase generation temperature<\/span>
                              \n<\/span>4.<\/b> Randomly select one version<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 2. Use version and effective-date metadata<\/b><\/p>\n

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

                              Version and effective-date metadata help a retrieval system distinguish current policies from historical versions. Without this information, semantic retrieval may return several versions that contain similar language but different requirements. The language model may then have difficulty determining which version should guide the answer. Metadata can be used to filter out superseded versions or prioritize the currently effective document. Maintaining this information during ingestion is therefore important for policy-heavy knowledge bases. When conflicts remain unresolved, the application should avoid presenting an uncertain answer as definitive and may need to communicate the relevant version information to the user.<\/span><\/p>\n

                              Question 355. What does retrieval recall measure in a RAG evaluation?<\/b><\/p>\n

                                \n
                              1. <\/b> How many relevant items were successfully retrieved from the available relevant items<\/span>
                                \n<\/span>2.<\/b> How many generated tokens were produced<\/span>
                                \n<\/span>3.<\/b> How quickly the model responded<\/span>
                                \n<\/span>4.<\/b> How many users accessed the system<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 1. How many relevant items were successfully retrieved from the available relevant items<\/b><\/p>\n

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

                                Retrieval recall measures the proportion of relevant documents or chunks that the retrieval system successfully returns from the set of relevant information available for a query. High recall means the system is less likely to miss useful evidence. This metric is important because the language model cannot use evidence that was never retrieved. However, maximizing recall alone can result in many irrelevant documents being returned, which can reduce precision and increase context size. Retrieval recall should therefore be considered alongside retrieval precision and downstream answer-quality metrics. Evaluation datasets with known relevant documents can help quantify retrieval performance consistently.<\/span><\/p>\n

                                Question 356. Why can document deduplication improve retrieval quality?<\/b><\/p>\n

                                  \n
                                1. <\/b> It reduces redundant results that may consume retrieval slots<\/span>
                                  \n<\/span>2.<\/b> It increases model temperature<\/span>
                                  \n<\/span>3.<\/b> It eliminates the need for chunking<\/span>
                                  \n<\/span>4.<\/b> It guarantees groundedness<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 1. It reduces redundant results that may consume retrieval slots<\/b><\/p>\n

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

                                  Duplicate documents can cause a retrieval system to return multiple copies of essentially the same information. This redundancy can consume top-k retrieval slots and reduce the diversity of evidence available to the language model. Deduplication can therefore improve the usefulness of the retrieved context by ensuring that different results provide distinct information where appropriate. Content hashes, source identifiers, canonical document IDs, and ingestion metadata can help identify duplicates. Deduplication is especially useful in enterprise knowledge bases where documents may be copied between repositories or published in multiple locations. It should be implemented carefully so that legitimately different versions are not incorrectly treated as duplicates.<\/span><\/p>\n

                                  Question 357. What is the primary role of production monitoring after deploying a RAG application?<\/b><\/p>\n

                                    \n
                                  1. <\/b> To observe performance, quality signals, and operational behavior over time<\/span>
                                    \n<\/span>2.<\/b> To automatically rewrite every user query<\/span>
                                    \n<\/span>3.<\/b> To eliminate evaluation<\/span>
                                    \n<\/span>4.<\/b> To replace document ingestion<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 1. To observe performance, quality signals, and operational behavior over time<\/b><\/p>\n

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

                                    Production monitoring provides visibility into how the application behaves under real workloads. Useful measurements may include latency, throughput, error rates, retrieval behavior, model usage, token consumption, cost, and user feedback. Monitoring can reveal problems that were not visible during controlled evaluation, such as changes in traffic patterns, data freshness issues, infrastructure bottlenecks, or unexpected retrieval behavior. These signals can feed into the application’s improvement cycle. Monitoring does not replace evaluation because production observations alone may not provide controlled comparisons between configurations. Instead, monitoring and evaluation complement each other by covering real-world behavior and controlled testing.<\/span><\/p>\n

                                    Question 358. Which approach is appropriate when a knowledge base changes frequently?<\/b><\/p>\n

                                      \n
                                    1. <\/b> Keep the retrieval index permanently unchanged<\/span>
                                      \n<\/span>2.<\/b> Use an ingestion process that detects and indexes relevant updates<\/span>
                                      \n<\/span>3.<\/b> Fine-tune the model after every document change<\/span>
                                      \n<\/span>4.<\/b> Remove document metadata<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 2. Use an ingestion process that detects and indexes relevant updates<\/b><\/p>\n

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

                                      Frequently changing knowledge requires a process that keeps the retrieval layer synchronized with authoritative source data. An ingestion pipeline can detect new, modified, or removed documents and update the relevant chunks, embeddings, metadata, and index entries. Incremental processing can reduce unnecessary work by focusing on documents that actually changed. This approach is often more practical than retraining or fine-tuning a language model every time factual information changes. The application should also preserve version and effective-date information when relevant. Monitoring ingestion freshness can help ensure that the retrieval system does not continue serving stale information after the source has been updated.<\/span><\/p>\n

                                      Question 359. What should be considered when selecting the top-k value for retrieval?<\/b><\/p>\n

                                        \n
                                      1. <\/b> The balance between retrieving sufficient evidence and limiting irrelevant context<\/span>
                                        \n<\/span>2.<\/b> Only the model’s name<\/span>
                                        \n<\/span>3.<\/b> The user’s screen size<\/span>
                                        \n<\/span>4.<\/b> Whether the application uses a dark interface<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 1. The balance between retrieving sufficient evidence and limiting irrelevant context<\/b><\/p>\n

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

                                        The top-k value determines how many retrieval candidates are selected for further processing. A small k may improve precision by limiting context but can miss useful evidence and reduce recall. A very large k can increase recall while also introducing irrelevant or redundant documents, increasing token usage and latency. The appropriate value therefore depends on the quality of the retriever, document collection, query characteristics, reranking strategy, and model context capacity. Teams should evaluate different values using representative queries rather than assuming that a larger or smaller number is always better. Top-k is one part of a broader retrieval-quality optimization process.<\/span><\/p>\n

                                        Question 360. Which workflow best supports reliable RAG application development?<\/b><\/p>\n

                                          \n
                                        1. <\/b> Change prompts and retrieval settings without testing<\/span>
                                          \n<\/span>2.<\/b> Deploy experimental configurations directly to production<\/span>
                                          \n<\/span>3.<\/b> Retrieve all documents and rely entirely on model judgment<\/span>
                                          \n<\/span>4.<\/b> Prepare quality data, evaluate retrieval and generation, monitor production, and iterate using versioned configurations<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 4. Prepare quality data, evaluate retrieval and generation, monitor production, and iterate using versioned configurations<\/b><\/p>\n

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

                                          Reliable RAG development requires an end-to-end lifecycle covering data quality, retrieval, generation, evaluation, deployment, and monitoring. The knowledge base should contain accurate and appropriately managed information, while ingestion should maintain current versions and useful metadata. Retrieval should be evaluated separately from final answer quality so that failures can be diagnosed accurately. Prompts, models, embeddings, and retrieval settings should be versioned to support reproducibility and rollback. After deployment, production monitoring and user feedback can reveal issues that controlled tests did not capture. This continuous, controlled process helps teams improve the application while maintaining traceability, reliability, and operational stability.<\/span><\/p>\n

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

                                          View Full Databricks Certified Generative AI Engineer Associate Exam Dumps and Practice Test Dumps   Question 341. What is the primary purpose of a system prompt in a generative AI application? To store document embeddings 2. To define high-level instructions and behavior for the model 3. To replace the retrieval index 4. To increase network […]<\/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\/18157"}],"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=18157"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/18157\/revisions"}],"predecessor-version":[{"id":18158,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/18157\/revisions\/18158"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=18157"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=18157"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=18157"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}