{"id":19160,"date":"2026-09-22T12:02:40","date_gmt":"2026-09-22T12:02:40","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=19160"},"modified":"2026-09-22T12:02:40","modified_gmt":"2026-09-22T12:02:40","slug":"pmi-cpmai-practice-test-questions-and-exam-dumps-part11-q201-220","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/pmi-cpmai-practice-test-questions-and-exam-dumps-part11-q201-220\/","title":{"rendered":"PMI CPMAI Practice Test Questions and Exam Dumps Part11 Q201-220"},"content":{"rendered":"

View Full <\/b>PMI CPMAI Exam Dumps<\/b><\/a> and Practice Test Dumps.<\/b><\/h2>\n

 <\/p>\n

Question 201<\/b><\/h3>\n

Which activity is most important when prioritizing multiple potential AI use cases?<\/b><\/p>\n

    \n
  1. Selecting the use case with the largest dataset<\/span><\/li>\n
  2. Choosing the use case with the most complex algorithm<\/span><\/li>\n
  3. Prioritizing use cases based on business value, feasibility, risk, and strategic alignment<\/span><\/li>\n
  4. Selecting the use case preferred by the technical team<\/span><\/li>\n<\/ol>\n

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

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

    AI use-case prioritization should consider more than technical complexity or dataset size. Organizations should evaluate expected business value, strategic alignment, technical feasibility, operational readiness, implementation cost, risk, data availability, and potential organizational impact. A use case with a highly sophisticated model may not be valuable if it does not address an important business need. Similarly, a valuable idea may need to be postponed if the required data or infrastructure is unavailable. A structured prioritization approach allows stakeholders to compare alternatives using consistent criteria. This helps organizations direct limited resources toward initiatives that have appropriate value while maintaining awareness of technical, operational, ethical, and business risks.<\/span><\/p>\n

    Question 202<\/b><\/h3>\n

    Which data-quality dimension focuses on whether data values correctly represent the real-world information they are intended to describe?<\/b><\/p>\n

      \n
    1. Accuracy<\/span><\/li>\n
    2. Timeliness<\/span><\/li>\n
    3. Completeness<\/span><\/li>\n
    4. Uniqueness<\/span><\/li>\n<\/ol>\n

      Correct Answer: 1<\/b><\/p>\n

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

      Accuracy refers to whether data correctly represents the real-world entities, events, or values it is intended to describe. For example, an incorrect customer address or an incorrectly recorded transaction amount represents an accuracy problem. Other dimensions measure different characteristics. Completeness concerns whether required values are present, timeliness concerns whether information is sufficiently current, and uniqueness concerns duplicate records or repeated representations of the same entity. Data-quality assessment should consider multiple dimensions because a dataset can be complete but inaccurate or accurate but outdated. Reliable AI systems require data-quality controls that address the dimensions most relevant to the intended use case and decision context.<\/span><\/p>\n

      Question 203<\/b><\/h3>\n

      What is the primary purpose of a requirements traceability matrix in an AI project?<\/b><\/p>\n

        \n
      1. To replace project planning<\/span><\/li>\n
      2. To connect requirements with corresponding deliverables, tests, and outcomes<\/span><\/li>\n
      3. To automatically train the AI model<\/span><\/li>\n
      4. To calculate the model’s inference latency<\/span><\/li>\n<\/ol>\n

        Correct Answer: 2<\/b><\/p>\n

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

        A requirements traceability matrix provides a structured way to connect project requirements with related deliverables, design elements, implementation activities, and verification or validation activities. In an AI project, this can help demonstrate that important business and technical requirements have been addressed and tested. Traceability also makes it easier to identify the impact of requirement changes and determine which components may need to be updated. The matrix does not train models or calculate inference latency. Instead, it supports project control, accountability, completeness, and change management. Maintaining traceability is especially useful when AI systems have multiple stakeholders and complex business, technical, security, or compliance requirements.<\/span><\/p>\n

        Question 204<\/b><\/h3>\n

        Which approach is most appropriate when an AI project has insufficient data for the intended model?<\/b><\/p>\n

          \n
        1. Ignore the limitation and proceed directly to production<\/span><\/li>\n
        2. Remove the business objective<\/span><\/li>\n
        3. Assess alternative data sources, data-generation approaches, or a revised solution scope<\/span><\/li>\n
        4. Assume that more model complexity will solve the problem<\/span><\/li>\n<\/ol>\n

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

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

          Insufficient data should be treated as a feasibility concern rather than ignored. The project team can investigate additional legitimate data sources, improve data collection processes, use suitable synthetic or augmented data where appropriate, or reconsider the scope of the AI solution. Domain experts and data owners should help determine whether available information is sufficiently representative and suitable for the intended use. Simply increasing model complexity cannot compensate for fundamentally inadequate data. If the data limitation cannot be resolved, the organization may need to modify the use case or postpone implementation. Early identification of data constraints helps prevent wasted development effort and reduces the risk of unreliable AI outcomes.<\/span><\/p>\n

          Question 205<\/b><\/h3>\n

          Which principle requires an AI system to have clearly identified individuals or groups responsible for its decisions and operation?<\/b><\/p>\n

            \n
          1. Accountability<\/span><\/li>\n
          2. Compression<\/span><\/li>\n
          3. Normalization<\/span><\/li>\n
          4. Clustering<\/span><\/li>\n<\/ol>\n

            Correct Answer: 1<\/b><\/p>\n

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

            Accountability means that responsibility for an AI system’s development, deployment, operation, and outcomes is clearly assigned. Organizations should identify appropriate owners and decision-makers for areas such as data management, model approval, security, monitoring, incident response, and business use. Accountability does not necessarily mean that one individual is responsible for every aspect of an AI system. Different roles may have different responsibilities throughout the lifecycle. Clear accountability helps ensure that important issues are addressed rather than left unresolved between technical and business teams. It also supports governance, auditability, and responsible decision-making by establishing who has authority to approve, modify, monitor, or retire an AI system.<\/span><\/p>\n

            Question 206<\/b><\/h3>\n

            Which technique is commonly used to reduce the dimensionality of numerical datasets while retaining important variation?<\/b><\/p>\n

              \n
            1. Tokenization<\/span><\/li>\n
            2. Principal Component Analysis<\/span><\/li>\n
            3. One-hot validation<\/span><\/li>\n
            4. Data encryption<\/span><\/li>\n<\/ol>\n

              Correct Answer: 2<\/b><\/p>\n

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

              Principal Component Analysis, or PCA, is a dimensionality-reduction technique that transforms correlated numerical variables into a smaller set of principal components. These components are constructed to capture important patterns of variation within the original data. Reducing dimensionality can simplify datasets, decrease computational requirements, and sometimes help address issues associated with highly correlated features. PCA should be applied carefully because transformed components may be less directly interpretable than the original variables. Tokenization is generally associated with processing text, encryption protects data, and validation is a model-development activity. PCA is therefore particularly useful when a numerical dataset contains many correlated features and dimensionality reduction is appropriate for the project.<\/span><\/p>\n

              Question 207<\/b><\/h3>\n

              What is the main purpose of establishing AI system ownership after deployment?<\/b><\/p>\n

                \n
              1. To ensure responsibility for ongoing operation, monitoring, maintenance, and decisions is clearly assigned<\/span><\/li>\n
              2. To prevent all future model updates<\/span><\/li>\n
              3. To eliminate the need for user training<\/span><\/li>\n
              4. To transfer every responsibility to the data-science team<\/span><\/li>\n<\/ol>\n

                Correct Answer: 1<\/b><\/p>\n

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

                AI system ownership establishes clear responsibility for the system after deployment. An owner or defined group may coordinate monitoring, maintenance, performance reviews, incident management, access decisions, changes, and retirement activities. Ownership should reflect the organization’s governance structure rather than assuming that every responsibility belongs to data scientists. Operational, security, compliance, business, and technical teams may each have defined roles. Clear ownership helps prevent situations where problems are identified but nobody has authority or responsibility to address them. It also supports lifecycle management by ensuring that the system continues to be monitored and maintained according to business objectives, technical requirements, risk controls, and governance expectations.<\/span><\/p>\n

                Question 208<\/b><\/h3>\n

                Which metric is particularly useful for evaluating a regression model when the magnitude of prediction errors should be penalized more strongly as errors become larger?<\/b><\/p>\n

                  \n
                1. Accuracy<\/span><\/li>\n
                2. Recall<\/span><\/li>\n
                3. Mean squared error<\/span><\/li>\n
                4. Precision<\/span><\/li>\n<\/ol>\n

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

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

                  Mean squared error, or MSE, calculates the average of the squared differences between predicted and actual values. Because errors are squared, larger errors receive disproportionately greater influence on the resulting metric. This can make MSE useful when large prediction errors are particularly undesirable. It is commonly used for evaluating regression models. Accuracy, precision, and recall are generally associated with classification tasks rather than continuous-value prediction. MSE should still be interpreted in the context of the business problem because its sensitivity to large errors may be either useful or undesirable depending on the application. Other regression measures, such as MAE or RMSE, may also provide complementary information.<\/span><\/p>\n

                  Question 209<\/b><\/h3>\n

                  Which action can help prevent data leakage during model development?<\/b><\/p>\n

                    \n
                  1. Using information that would only become available after the prediction event<\/span><\/li>\n
                  2. Separating training, validation, and test data appropriately and excluding future or target-derived information<\/span><\/li>\n
                  3. Increasing the number of model parameters<\/span><\/li>\n
                  4. Removing all validation procedures<\/span><\/li>\n<\/ol>\n

                    Correct Answer: 2<\/b><\/p>\n

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

                    Data leakage occurs when information that should not be available during prediction is unintentionally used during model training or evaluation. This can produce unrealistically strong performance estimates and cause poor results after deployment. Teams can reduce leakage by carefully separating datasets, reviewing feature-generation processes, ensuring that future information is excluded, and preventing target-derived information from entering inappropriate features. Temporal problems require particular attention because data collected after the prediction point may accidentally influence training. Increasing model complexity does not prevent leakage. Careful data preparation, feature review, pipeline controls, and independent validation are important safeguards for producing realistic estimates of model performance.<\/span><\/p>\n

                    Question 210<\/b><\/h3>\n

                    Which security control is most appropriate for limiting AI system access according to a user’s responsibilities?<\/b><\/p>\n

                      \n
                    1. Role-based access control<\/span><\/li>\n
                    2. Random data deletion<\/span><\/li>\n
                    3. Model retraining<\/span><\/li>\n
                    4. Feature scaling<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 1<\/b><\/p>\n

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

                      Role-based access control, or RBAC, assigns permissions according to defined roles and responsibilities. For an AI system, different roles may require different access levels to training data, model artifacts, deployment environments, monitoring dashboards, or administrative functions. RBAC supports the principle of least privilege by limiting users to the permissions necessary for their responsibilities. Random data deletion and feature scaling are unrelated to access management, while model retraining addresses model lifecycle activities. RBAC should be combined with authentication, logging, periodic access reviews, and other security controls. Proper access management reduces the likelihood of unauthorized changes, inappropriate data access, and accidental modification of critical AI resources.<\/span><\/p>\n

                      Question 211<\/b><\/h3>\n

                      What is the primary purpose of a proof of concept in an AI initiative?<\/b><\/p>\n

                        \n
                      1. To demonstrate whether a proposed approach is technically and practically viable on a limited scale<\/span><\/li>\n
                      2. To complete full production deployment immediately<\/span><\/li>\n
                      3. To replace business requirements<\/span><\/li>\n
                      4. To guarantee long-term financial benefits<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 1<\/b><\/p>\n

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

                        A proof of concept, or PoC, is a limited experiment used to determine whether a proposed AI approach can work under relevant conditions. It may examine data availability, model feasibility, integration challenges, performance, or other critical assumptions before significant resources are committed to full implementation. A PoC is not the same as production deployment and does not guarantee long-term financial benefits. Its purpose is to generate evidence about important feasibility questions and reduce uncertainty. Clear success criteria should be established before the PoC begins so stakeholders can determine whether the evidence supports continuing, modifying, or stopping the proposed initiative.<\/span><\/p>\n

                        Question 212<\/b><\/h3>\n

                        Which practice best supports reproducibility when multiple teams develop AI models?<\/b><\/p>\n

                          \n
                        1. Allowing each team to use undocumented datasets<\/span><\/li>\n
                        2. Tracking code, data versions, configurations, dependencies, and experiment results<\/span><\/li>\n
                        3. Changing model settings without recording them<\/span><\/li>\n
                        4. Deleting previous model versions after every release<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 2<\/b><\/p>\n

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

                          Reproducibility requires enough information to recreate or meaningfully reproduce an experiment, model, or result. AI teams should track relevant code versions, dataset versions, configurations, dependencies, model parameters, experiment results, and other important environmental details. Version control and experiment-tracking systems can help maintain this information consistently. Undocumented datasets and unrecorded configuration changes make it difficult to understand how a result was produced. Deleting previous versions also reduces the ability to investigate historical behavior. Reproducibility supports reliable development, auditing, troubleshooting, collaboration, and validation because teams can determine what changed and reproduce important results when necessary.<\/span><\/p>\n

                          Question 213<\/b><\/h3>\n

                          Which type of AI risk is most directly associated with exposing confidential customer information through model outputs?<\/b><\/p>\n

                            \n
                          1. Scalability risk<\/span><\/li>\n
                          2. Privacy risk<\/span><\/li>\n
                          3. Latency risk<\/span><\/li>\n
                          4. Availability risk<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 2<\/b><\/p>\n

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

                            Exposure of confidential customer information through AI outputs is primarily a privacy risk because sensitive or personal information may be disclosed to unauthorized users or contexts. Privacy risks can arise from training data, model memorization, poorly designed prompts, excessive data access, insecure integrations, or inappropriate output handling. Organizations should assess what information the system can access and potentially reproduce, then implement controls such as data minimization, access restrictions, output filtering, privacy testing, and appropriate retention practices. Scalability concerns system capacity, latency concerns response time, and availability concerns whether the service remains accessible. Proper privacy assessment should occur before and after deployment because risks can change during operational use.<\/span><\/p>\n

                            Question 214<\/b><\/h3>\n

                            Why should an AI project define acceptance criteria before final implementation?<\/b><\/p>\n

                              \n
                            1. To establish objective conditions that determine whether the delivered solution meets agreed requirements<\/span><\/li>\n
                            2. To guarantee that no stakeholder will request changes<\/span><\/li>\n
                            3. To eliminate the need for model testing<\/span><\/li>\n
                            4. To select the programming language automatically<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 1<\/b><\/p>\n

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

                              Acceptance criteria define the conditions that must be satisfied for stakeholders to consider an AI deliverable acceptable. These criteria may address functional behavior, model performance, business outcomes, security requirements, usability, integration, reliability, or other agreed requirements. Defining them before final implementation helps create objective expectations and reduces ambiguity during acceptance. Acceptance criteria do not prevent future changes and do not replace technical or security testing. Instead, they provide a reference for determining whether the delivered solution satisfies agreed expectations. Clear criteria also improve communication between business stakeholders, project managers, technical teams, and users by making success conditions explicit and measurable.<\/span><\/p>\n

                              Question 215<\/b><\/h3>\n

                              Which approach is most appropriate for evaluating fairness when an AI model produces different outcomes across demographic or other relevant groups?<\/b><\/p>\n

                                \n
                              1. Evaluate performance and outcomes separately across relevant groups<\/span><\/li>\n
                              2. Evaluate only the overall accuracy<\/span><\/li>\n
                              3. Remove all group information without further analysis<\/span><\/li>\n
                              4. Assume that identical model code guarantees identical outcomes<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 1<\/b><\/p>\n

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

                                Fairness assessment often requires examining model performance and outcomes across relevant groups rather than relying only on an overall metric. Teams may compare measures such as error rates, false-positive rates, false-negative rates, precision, recall, or other outcome measures depending on the use case. Group-level evaluation can reveal disparities that aggregate performance may conceal. Removing group information from an analysis may make it harder to identify such disparities, and identical model code does not guarantee identical outcomes across groups because data distributions and historical patterns can differ. Fairness analysis should therefore be aligned with the use case, applicable requirements, stakeholder concerns, and appropriate ethical and governance considerations.<\/span><\/p>\n

                                Question 216<\/b><\/h3>\n

                                What is the main purpose of a model registry in an AI lifecycle?<\/b><\/p>\n

                                  \n
                                1. To store and manage model versions and associated metadata for controlled lifecycle management<\/span><\/li>\n
                                2. To replace all data storage systems<\/span><\/li>\n
                                3. To automatically eliminate model bias<\/span><\/li>\n
                                4. To provide internet access to every model<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 1<\/b><\/p>\n

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

                                  A model registry provides a centralized mechanism for managing model versions and associated metadata throughout the AI lifecycle. Information may include model identifiers, versions, training details, evaluation results, approval status, deployment status, ownership, and relevant documentation. This supports controlled promotion of models between development, testing, and production environments. A model registry does not automatically eliminate bias or replace data storage systems. Its main value is lifecycle governance and traceability. By maintaining a clear record of which model versions exist and where they are approved or deployed, organizations can improve reproducibility, auditing, rollback capabilities, and operational control.<\/span><\/p>\n

                                  Question 217<\/b><\/h3>\n

                                  Which situation is an example of concept drift?<\/b><\/p>\n

                                    \n
                                  1. The model’s input data format changes from CSV to JSON<\/span><\/li>\n
                                  2. The relationship between input features and the target outcome changes over time<\/span><\/li>\n
                                  3. The model file is moved to a different server<\/span><\/li>\n
                                  4. A user changes their password<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 2<\/b><\/p>\n

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

                                    Concept drift occurs when the underlying relationship between input variables and the target outcome changes over time. For example, customer behavior may change so that patterns that previously predicted purchasing behavior no longer have the same relationship with the outcome. This can cause model performance to decline even if the general structure of incoming data appears similar. Concept drift differs from data drift, where the distribution of input data changes. Changes in file format or server location are operational or technical changes rather than concept drift. Monitoring should therefore consider both data characteristics and actual model performance to determine whether changing real-world relationships require investigation or retraining.<\/span><\/p>\n

                                    Question 218<\/b><\/h3>\n

                                    Which approach can help an organization evaluate whether a third-party AI vendor is suitable for a high-risk use case?<\/b><\/p>\n

                                      \n
                                    1. Evaluate the vendor’s security, privacy, performance, governance, documentation, and contractual controls<\/span><\/li>\n
                                    2. Select the vendor solely because its model has the largest parameter count<\/span><\/li>\n
                                    3. Avoid reviewing vendor documentation<\/span><\/li>\n
                                    4. Allow the vendor unrestricted access to all organizational data<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 1<\/b><\/p>\n

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

                                      Third-party AI evaluation should consider multiple dimensions relevant to the organization’s risk and requirements. These can include model performance, security controls, privacy practices, data handling, documentation, service reliability, incident response, governance, regulatory obligations, support arrangements, and contractual terms. For high-risk use cases, organizations should understand how the vendor manages data, model changes, access, outages, and security incidents. Model size alone does not establish suitability. Unrestricted data access can introduce unnecessary exposure. A structured vendor assessment helps organizations identify risks before adoption and establish appropriate contractual and operational controls for the AI service throughout its lifecycle.<\/span><\/p>\n

                                      Question 219<\/b><\/h3>\n

                                      Which deployment method gradually exposes a new model to a limited portion of production traffic before broader rollout?<\/b><\/p>\n

                                        \n
                                      1. Shadow deployment<\/span><\/li>\n
                                      2. Canary deployment<\/span><\/li>\n
                                      3. Offline training<\/span><\/li>\n
                                      4. Batch preprocessing<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 2<\/b><\/p>\n

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

                                        Canary deployment introduces a new model to a limited portion of production traffic while the existing system continues serving most users. Teams can monitor performance, errors, latency, resource consumption, and business outcomes before increasing the new model’s traffic allocation. This approach can reduce deployment risk because problems can be detected before the new model is exposed broadly. Shadow deployment is different because the new model may process production inputs without making the operational decision. Offline training and batch preprocessing are development or data-processing activities rather than controlled production rollout strategies. Canary deployment is therefore useful when teams want to validate a model incrementally under real operating conditions.<\/span><\/p>\n

                                        Question 220<\/b><\/h3>\n

                                        Which activity is most useful for determining whether an AI system should be retrained after deployment?<\/b><\/p>\n

                                          \n
                                        1. Monitoring relevant performance, data, and drift indicators against defined thresholds<\/span><\/li>\n
                                        2. Retraining the model on a fixed monthly schedule regardless of performance<\/span><\/li>\n
                                        3. Ignoring changes in incoming data<\/span><\/li>\n
                                        4. Removing the model’s historical performance records<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 1<\/b><\/p>\n

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

                                          Retraining decisions should be supported by evidence from ongoing monitoring. Teams can track model performance, input-data distributions, concept drift indicators, business outcomes, error patterns, and other relevant signals against predefined thresholds or decision criteria. If monitoring shows meaningful degradation or changes in the operating environment, the team can investigate whether retraining or another corrective action is appropriate. A fixed schedule may sometimes be useful, but retraining without considering actual system conditions can waste resources or introduce unnecessary changes. Historical performance records should be retained because they provide valuable evidence for comparing versions and determining whether a retraining intervention actually improves the system.<\/span><\/p>\n

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

                                          View Full PMI CPMAI Exam Dumps and Practice Test Dumps.   Question 201 Which activity is most important when prioritizing multiple potential AI use cases? Selecting the use case with the largest dataset Choosing the use case with the most complex algorithm Prioritizing use cases based on business value, feasibility, risk, and strategic alignment Selecting […]<\/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\/19160"}],"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=19160"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/19160\/revisions"}],"predecessor-version":[{"id":19161,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/19160\/revisions\/19161"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=19160"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=19160"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=19160"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}