{"id":22415,"date":"2026-09-25T12:42:27","date_gmt":"2026-09-25T12:42:27","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=22415"},"modified":"2026-09-25T12:42:27","modified_gmt":"2026-09-25T12:42:27","slug":"microsoft-ai-102-practice-test-questions-and-exam-dumps-part15-q281-300","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/microsoft-ai-102-practice-test-questions-and-exam-dumps-part15-q281-300\/","title":{"rendered":"Microsoft AI-102 Practice Test Questions and Exam Dumps Part15 Q281-300"},"content":{"rendered":"

View Full <\/b>Microsoft AI-102 Exam Dumps<\/b><\/a> and Practice Test Dumps.<\/b><\/h2>\n

 <\/p>\n

Question 281<\/b><\/h3>\n

Which Microsoft Foundry capability helps organize models, tools, and AI application resources within a collaborative workspace?<\/b><\/p>\n

    \n
  1. Speech profile<\/span><\/li>\n
  2. Project<\/span><\/li>\n
  3. OCR<\/span><\/li>\n
  4. Facet<\/span><\/li>\n<\/ol>\n

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

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

    A Microsoft Foundry project provides a workspace for organizing resources used during AI application development. Depending on the solution, developers can work with models, evaluations, connections, tools, and other components within the project context. This organization can make collaboration and application lifecycle management easier because related resources are associated with a common development environment. Projects can support workflows involving experimentation, evaluation, and deployment. Developers should still apply appropriate access controls to project resources and connections because a project workspace does not automatically mean that every user should have unrestricted access to every underlying model, tool, or data source.<\/span><\/p>\n

    Question 282<\/b><\/h3>\n

    A company wants to identify customer requests such as refunds, cancellations, and product inquiries from incoming messages. Which capability is appropriate when these categories are organization-specific?<\/b><\/p>\n

      \n
    1. OCR<\/span><\/li>\n
    2. Image analysis<\/span><\/li>\n
    3. Text-to-speech<\/span><\/li>\n
    4. Custom text classification<\/span><\/li>\n<\/ol>\n

      Correct Answer: 4<\/b><\/p>\n

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

      Custom text classification is appropriate when an organization needs to classify text according to business-specific categories. Customer requests such as refunds, cancellations, complaints, and product inquiries can be represented as labeled classes using example messages. The model learns patterns associated with those categories and can classify new messages. Training data should contain enough variation to represent different ways customers express the same intent. Developers should maintain separate evaluation examples and monitor performance after deployment. If categories overlap substantially, their definitions should be clarified because ambiguous labels can reduce classification quality and make automated routing less reliable.<\/span><\/p>\n

      Question 283<\/b><\/h3>\n

      Which Azure AI Search option can use a graph-based approximate nearest-neighbor method for vector retrieval?<\/b><\/p>\n

        \n
      1. HNSW<\/span><\/li>\n
      2. Faceting<\/span><\/li>\n
      3. BM25<\/span><\/li>\n
      4. Exhaustive KNN<\/span><\/li>\n<\/ol>\n

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

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

        HNSW, or Hierarchical Navigable Small World, is a graph-based algorithm used for approximate nearest-neighbor vector search. It is designed to efficiently identify vectors that are similar to a query without comparing the query against every stored vector. This can provide strong retrieval performance for large vector collections while reducing the computational work required for each query. Exhaustive KNN instead performs exact nearest-neighbor comparisons. Developers should choose vector-search settings based on factors such as dataset size, latency targets, resource usage, and acceptable retrieval trade-offs. Testing with representative queries is important before selecting production settings.<\/span><\/p>\n

        Question 284<\/b><\/h3>\n

        A developer wants an AI application to retrieve only documents belonging to the user’s assigned region. Which search technique should be applied?<\/b><\/p>\n

          \n
        1. Speech synthesis<\/span><\/li>\n
        2. Metadata filtering<\/span><\/li>\n
        3. Image captioning<\/span><\/li>\n
        4. OCR<\/span><\/li>\n<\/ol>\n

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

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

          Metadata filtering can restrict search results according to structured information associated with documents. If documents contain a region field, the application can use that field to retrieve only content corresponding to the user’s authorized region. This approach can be combined with keyword, vector, hybrid, or semantic retrieval. In security-sensitive applications, filtering should be designed as part of a broader authorization strategy rather than being the only protection mechanism. Developers should ensure that users cannot manipulate filter values to bypass access restrictions. The underlying identity and authorization system should independently determine which data each user is permitted to access.<\/span><\/p>\n

          Question 285<\/b><\/h3>\n

          Which Azure AI Speech feature is useful when an application needs to recognize specialized vocabulary such as medical or technical terminology?<\/b><\/p>\n

            \n
          1. Document extraction<\/span><\/li>\n
          2. Semantic ranking<\/span><\/li>\n
          3. Custom speech<\/span><\/li>\n
          4. Image classification<\/span><\/li>\n<\/ol>\n

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

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

            Custom speech can help improve speech recognition for specialized terminology, domain-specific vocabulary, or other language patterns that may not be handled optimally by a general speech model. This can be useful in industries where users frequently mention technical names, product terms, medical terminology, or organization-specific language. Developers should provide representative data and evaluate recognition performance using realistic speakers and environments. Customization should focus on terminology and speech patterns that matter to the application. Even after customization, testing should continue with production-like audio because background noise, accents, microphone quality, and speaking styles can influence recognition accuracy.<\/span><\/p>\n

            Question 286<\/b><\/h3>\n

            A RAG application retrieves documents that are relevant but too large to fit efficiently into the model context. What should the developer consider?<\/b><\/p>\n

              \n
            1. Chunking<\/span><\/li>\n
            2. Faceting<\/span><\/li>\n
            3. Image resizing<\/span><\/li>\n
            4. Speech translation<\/span><\/li>\n<\/ol>\n

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

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

              Chunking divides large documents into smaller sections that can be retrieved independently. This can improve RAG performance because the search system can return focused pieces of information rather than entire documents containing large amounts of unrelated content. Effective chunking should preserve enough surrounding context to make each section meaningful while avoiding unnecessarily large retrieval units. Developers should consider document structure, headings, paragraphs, tables, and semantic boundaries when designing chunks. Chunk size should be evaluated together with embedding quality and retrieval performance. Poor chunking can cause important context to be separated or can produce retrieval results that are too broad.<\/span><\/p>\n

              Question 287<\/b><\/h3>\n

              Which Azure AI Search capability can provide a concise passage from a result that is relevant to a natural-language query?<\/b><\/p>\n

                \n
              1. Indexer<\/span><\/li>\n
              2. HNSW<\/span><\/li>\n
              3. Semantic captions<\/span><\/li>\n
              4. Faceting<\/span><\/li>\n<\/ol>\n

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

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

                Semantic captions can provide concise, relevant extracts from search results when semantic search is used. These captions can help users quickly understand why a document matches their natural-language query without reading the entire result. They are generated from relevant content within the indexed document and can improve the usability of search interfaces. Semantic captions are different from ordinary snippets generated through simple keyword matching because semantic search uses language understanding to identify useful passages. Developers should ensure that the indexed content and semantic configuration contain meaningful textual fields so that captions can provide useful information.<\/span><\/p>\n

                Question 288<\/b><\/h3>\n

                An AI assistant must call a service to check the status of a customer’s shipment. Which design best represents an agent tool?<\/b><\/p>\n

                  \n
                1. Search facet<\/span><\/li>\n
                2. OCR field<\/span><\/li>\n
                3. Image embedding<\/span><\/li>\n
                4. API operation<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 4<\/b><\/p>\n

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

                  An API operation can be exposed to an AI agent as a controlled tool that performs a specific business function, such as checking shipment status. The agent can determine when the tool is relevant, provide required parameters, and use the returned information to formulate a response. The application should validate parameters and enforce authorization before executing the operation. Tool permissions should be limited to the actions the agent actually requires. Developers should also define clear error handling for unavailable services, invalid shipment identifiers, or missing data. This architecture allows current information to be retrieved without requiring the model itself to contain that information.<\/span><\/p>\n

                  Question 289<\/b><\/h3>\n

                  Which Azure AI Document Intelligence capability can classify documents before routing them to specialized extraction models?<\/b><\/p>\n

                    \n
                  1. Custom classification<\/span><\/li>\n
                  2. Vector ranking<\/span><\/li>\n
                  3. Text-to-speech<\/span><\/li>\n
                  4. Sentiment analysis<\/span><\/li>\n<\/ol>\n

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

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

                    Document Intelligence custom classification can identify document types according to categories defined by an organization. This can be useful when an incoming collection contains multiple document types and each type requires different extraction processing. For example, an application could classify documents as invoices, purchase orders, contracts, or application forms before sending them to the appropriate downstream model. Training examples should represent the layouts and variations expected in production. Classification should be evaluated using documents not included in training. A well-designed routing workflow can reduce manual processing and help ensure that documents reach the correct extraction model.<\/span><\/p>\n

                    Question 290<\/b><\/h3>\n

                    Which prompt design practice helps reduce ambiguity when asking a generative AI model to perform a specific task?<\/b><\/p>\n

                      \n
                    1. Conflicting requirements<\/span><\/li>\n
                    2. Clear instructions<\/span><\/li>\n
                    3. Missing context<\/span><\/li>\n
                    4. Random wording<\/span><\/li>\n<\/ol>\n

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

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

                      Clear instructions help a generative AI model understand the task, expected output, constraints, and relevant context. Developers should specify important requirements explicitly rather than assuming the model will infer them correctly. Depending on the task, instructions can define the desired format, audience, scope, tone, or rules for handling missing information. Clear prompting should still be combined with evaluation because an instruction that appears unambiguous may behave differently across models or inputs. Developers should test prompts using representative examples and include difficult cases to determine whether the instructions consistently produce the intended behavior.<\/span><\/p>\n

                      Question 291<\/b><\/h3>\n

                      A search application must support both filtering by document type and vector similarity retrieval. What should the index contain?<\/b><\/p>\n

                        \n
                      1. Audio fields only<\/span><\/li>\n
                      2. Image captions only<\/span><\/li>\n
                      3. Speech profiles<\/span><\/li>\n
                      4. Filterable metadata and vector fields<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 4<\/b><\/p>\n

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

                        The search index should contain structured metadata fields configured for filtering and vector fields configured for vector retrieval. For example, a document type field can be marked filterable so queries can restrict results to invoices, policies, manuals, or other categories. The document’s embedding can be stored in a vector field for semantic similarity retrieval. These capabilities can work together in the same query workflow, allowing an application to combine semantic retrieval with metadata restrictions. Developers should design the schema carefully and verify that vector dimensions, field attributes, and filterable metadata correspond to actual application requirements.<\/span><\/p>\n

                        Question 292<\/b><\/h3>\n

                        Which Microsoft Foundry capability can help determine whether a prompt change improves an AI application’s responses?<\/b><\/p>\n

                          \n
                        1. OCR<\/span><\/li>\n
                        2. Speech recognition<\/span><\/li>\n
                        3. Evaluation<\/span><\/li>\n
                        4. Image cropping<\/span><\/li>\n<\/ol>\n

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

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

                          Evaluation allows developers to compare AI application behavior using defined test inputs and criteria. When a prompt changes, the same evaluation dataset can be used to determine whether response quality improves, declines, or changes in specific areas. Depending on the application, useful measures can include relevance, groundedness, coherence, safety, or task-specific success. Consistent evaluation reduces reliance on subjective impressions from a few examples. Developers should also consider latency and cost because a prompt that improves one quality metric may increase resource consumption. Evaluation should be repeated whenever significant changes are made to prompts, models, tools, or retrieval pipelines.<\/span><\/p>\n

                          Question 293<\/b><\/h3>\n

                          A developer needs to identify relationships between a user’s query and relevant passages in a knowledge base using vector similarity. Which data representation is required?<\/b><\/p>\n

                            \n
                          1. Embeddings<\/span><\/li>\n
                          2. Bounding boxes<\/span><\/li>\n
                          3. OCR images<\/span><\/li>\n
                          4. Audio waveforms<\/span><\/li>\n<\/ol>\n

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

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

                            Embeddings represent text or other supported content as numerical vectors that capture semantic characteristics. When both a user query and knowledge-base passages are represented as embeddings, a vector search system can compare their positions in the vector space and retrieve content with similar meaning. This allows an application to find relevant passages even when the wording differs. Embeddings are commonly used in RAG architectures and semantic search systems. Developers should select an appropriate embedding model and ensure that generated vectors match the index field’s expected dimensions. Retrieval quality should then be measured with realistic questions and relevant source documents.<\/span><\/p>\n

                            Question 294<\/b><\/h3>\n

                            Which Azure AI Language capability can identify phrases that represent important concepts without assigning a custom category?<\/b><\/p>\n

                              \n
                            1. Pronunciation assessment<\/span><\/li>\n
                            2. Speech translation<\/span><\/li>\n
                            3. Key phrase extraction<\/span><\/li>\n
                            4. Custom classification<\/span><\/li>\n<\/ol>\n

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

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

                              Key phrase extraction identifies important concepts or phrases within unstructured text. It can help summarize the main topics present in customer comments, reports, surveys, or other collections of documents. The extracted phrases can be used for search, indexing, analytics, content organization, or additional processing. Unlike custom classification, key phrase extraction does not require the developer to define business categories for every input. It also differs from named entity recognition, which identifies entities such as people, organizations, or locations. Developers should test the output on representative text because phrase importance can depend on wording, context, and document length.<\/span><\/p>\n

                              Question 295<\/b><\/h3>\n

                              Which approach helps an AI application avoid exposing sensitive information through an overly broad tool permission?<\/b><\/p>\n

                                \n
                              1. Full administrative access<\/span><\/li>\n
                              2. Shared credentials<\/span><\/li>\n
                              3. Anonymous execution<\/span><\/li>\n
                              4. Least privilege<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 4<\/b><\/p>\n

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

                                Least privilege limits a tool, application identity, or agent to only the permissions necessary for its intended task. This is particularly important when AI agents can invoke tools that access business systems because excessive permissions can increase the consequences of an incorrect or malicious action. For example, a reporting agent that only needs read access should not receive permission to delete records. Developers should configure authorization carefully, separate sensitive operations, validate tool inputs, and review permissions periodically. Least privilege should be combined with authentication, monitoring, logging, and appropriate user confirmation for high-impact operations.<\/span><\/p>\n

                                Question 296<\/b><\/h3>\n

                                An application needs to process a large set of PDFs and extract entities and key phrases before indexing them. Which architecture is suitable?<\/b><\/p>\n

                                  \n
                                1. AI Search skillset pipeline<\/span><\/li>\n
                                2. Manual indexing only<\/span><\/li>\n
                                3. Image classification pipeline<\/span><\/li>\n
                                4. Speech-only pipeline<\/span><\/li>\n<\/ol>\n

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

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

                                  An Azure AI Search skillset pipeline can enrich documents during indexing by applying supported AI skills to extracted content. A workflow can use OCR when necessary and then apply language-processing capabilities such as entity recognition or key phrase extraction before mapping enriched information into the search index. This approach is useful for knowledge-mining scenarios where raw documents need additional processing before users search them. Developers should configure the data source, indexer, skillset, field mappings, and target index consistently. The enrichment pipeline should also be evaluated to confirm that extracted information improves search usefulness rather than introducing unnecessary or inaccurate fields.<\/span><\/p>\n

                                  Question 297<\/b><\/h3>\n

                                  A generative AI assistant should answer questions using only information returned by an approved knowledge source. Which instruction is most relevant?<\/b><\/p>\n

                                    \n
                                  1. Generate unsupported details<\/span><\/li>\n
                                  2. Ground responses in retrieved context<\/span><\/li>\n
                                  3. Ignore retrieved documents<\/span><\/li>\n
                                  4. Use unrelated examples<\/span><\/li>\n<\/ol>\n

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

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

                                    In a grounded RAG workflow, instructions can tell the model to base its response on the retrieved context and avoid introducing unsupported information. This helps align the generated answer with the approved knowledge source. However, instructions alone do not guarantee grounded responses. Developers should also evaluate retrieval quality, monitor outputs, use appropriate safety controls, and define behavior for cases where the retrieved context does not contain an answer. A useful application may explicitly instruct the model to acknowledge insufficient evidence instead of inventing information. Groundedness testing can then measure whether generated responses remain supported by retrieved content.<\/span><\/p>\n

                                    Question 298<\/b><\/h3>\n

                                    Which Azure AI Search feature can influence how text search results are scored using field-specific weighting?<\/b><\/p>\n

                                      \n
                                    1. Skillset<\/span><\/li>\n
                                    2. Vector field<\/span><\/li>\n
                                    3. Scoring profile<\/span><\/li>\n
                                    4. Indexer<\/span><\/li>\n<\/ol>\n

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

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

                                      A scoring profile can influence the ranking of search results by applying customized scoring rules to fields or other supported factors. This can be useful when certain fields should contribute more strongly to relevance than others. For example, an application may want matches in a product name field to have greater importance than matches in a long descriptive field. Scoring profiles are different from semantic ranking and vector similarity, although they can be part of a broader search strategy. Developers should test scoring changes with representative queries because modifying ranking behavior can improve some searches while producing less desirable ordering for others.<\/span><\/p>\n

                                      Question 299<\/b><\/h3>\n

                                      Which Azure AI Vision scenario requires identifying the position and size of detected objects rather than simply assigning an image category?<\/b><\/p>\n

                                        \n
                                      1. Image classification<\/span><\/li>\n
                                      2. Sentiment analysis<\/span><\/li>\n
                                      3. Object detection<\/span><\/li>\n
                                      4. Image captioning<\/span><\/li>\n<\/ol>\n

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

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

                                        Object detection is used when an application needs to identify individual objects and determine where they occur within an image. The result can include object categories and corresponding locations or bounding regions. This differs from image classification, which generally assigns one or more categories to the image without providing the locations of individual objects. Image captioning instead generates a natural-language description of visual content. Developers should choose object detection when downstream logic depends on spatial information, such as counting items, monitoring locations, or identifying objects within specific regions. Performance should be tested using realistic images and environmental conditions.<\/span><\/p>\n

                                        Question 300<\/b><\/h3>\n

                                        A team changes its AI model and wants to verify that the new version has not introduced unexpected quality regressions. What should it do?<\/b><\/p>\n

                                          \n
                                        1. Test only one prompt<\/span><\/li>\n
                                        2. Delete previous results<\/span><\/li>\n
                                        3. Skip comparison<\/span><\/li>\n
                                        4. Run the existing evaluation set<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 4<\/b><\/p>\n

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

                                          Running the existing evaluation set allows a team to compare the new model with previous results using consistent inputs and criteria. This can reveal regressions in areas such as relevance, correctness, groundedness, safety, or task performance. A stable evaluation dataset provides a baseline against which changes can be measured. Developers should examine both improvements and degradations because a new model may perform better for some scenarios while becoming weaker in others. Evaluation should be supplemented with production-like testing, especially when the model change affects tool use, retrieval, response formatting, latency, or cost.<\/span><\/p>\n

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

                                          View Full Microsoft AI-102 Exam Dumps and Practice Test Dumps.   Question 281 Which Microsoft Foundry capability helps organize models, tools, and AI application resources within a collaborative workspace? Speech profile Project OCR Facet Correct Answer: 2 Explanation A Microsoft Foundry project provides a workspace for organizing resources used during AI application development. Depending on […]<\/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\/22415"}],"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=22415"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/22415\/revisions"}],"predecessor-version":[{"id":22416,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/22415\/revisions\/22416"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=22415"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=22415"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=22415"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}