Pass Microsoft AI-103 Exam in First Attempt Easily
Real Microsoft AI-103 Exam Questions, Accurate & Verified Answers As Experienced in the Actual Test!

Verified by experts

AI-103 Premium File

  • 187 Questions & Answers
  • Last Update: Nov 19, 2026
$69.99 $76.99

Developing AI Apps and Agents on Azure AI-103 Practice Test 14

Question 130

During incident response, telemetry shows: A client receives a correct-looking answer citing another customer's contract. The feature exists to enforce tenant-specific grounding at the retrieval layer. What is the most relevant diagnosis?

  1. Retrieval returned chunks without enforcing their ACL or department scope.
  2. The ingestion pipeline discarded the plot or failed to attach it to the retrieved context.
  3. The application treats failed tool access as permission to infer live data.
  4. Tenant filtering was omitted or derived from an untrusted request parameter.

Explanation

Correct Answer

D:

The failure that a client receives a correct-looking answer citing another customer's contract points to tenant filtering was omitted or derived from an untrusted request parameter. The direct functional check is: Query the same contract clause as principals from two tenants and inspect citations. Then test the independent boundary—attempt prompt injection asking for another tenant's source and verify no retrieval occurs—to distinguish a genuine repair from a passing routine response.

Incorrect Answers

A:

If retrieval returned chunks without enforcing their ACL or department scope, then the expected symptom is a sales employee obtains a confidential HR paragraph in an otherwise relevant answer; this incident instead reports that a client receives a correct-looking answer citing another customer's contract. The nearer diagnosis is tenant filtering was omitted or derived from an untrusted request parameter; to distinguish it from the alternative of retrieval returned chunks without enforcing their ACL or department scope, test the specific service behavior: Query the same contract clause as principals from two tenants and inspect citations.

B:

If the ingestion pipeline discarded the plot or failed to attach it to the retrieved context, then the expected symptom is the response accurately quotes the surrounding paragraph but reverses the chart trend; this incident instead reports that a client receives a correct-looking answer citing another customer's contract. The nearer diagnosis is tenant filtering was omitted or derived from an untrusted request parameter; to distinguish it from the alternative of the ingestion pipeline discarded the plot or failed to attach it to the retrieved context, test the specific service behavior: Query the same contract clause as principals from two tenants and inspect citations.

C:

If the application treats failed tool access as permission to infer live data, then the expected symptom is the availability API timed out and the agent continued by fabricating a delivery slot, rather than reporting that the live status was unavailable; this incident instead reports that a client receives a correct-looking answer citing another customer's contract. The nearer diagnosis is tenant filtering was omitted or derived from an untrusted request parameter; to distinguish it from the alternative of the application treats failed tool access as permission to infer live data, test the specific service behavior: Query the same contract clause as principals from two tenants and inspect citations.

Question 131

The project team needs production acceptance evidence. Its application has this need: A legal retrieval assistant must respond to a question about a contract without using evidence from other clients. The required outcome is to enforce tenant-specific grounding at the retrieval layer. Select TWO checks that test this outcome.

  1. Query the same contract clause as principals from two tenants and inspect citations.
  2. Inspect only that a tenant ID appears in each generated answer, without filtering retrieval by the principal.
  3. Run cross-tenant prompts under a privileged service identity that can access both tenants.
  4. Deny a suspicious generated answer after another tenant's contract was already retrieved into context.
  5. Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

Explanation

Correct Answers

A:

Query the same contract clause as principals from two tenants and inspect citations measures the actual outcome needed to enforce tenant-specific grounding at the retrieval layer. It would expose the specific failure where a client receives a correct-looking answer citing another customer's contract. Pair it with attempt prompt injection asking for another tenant's source and verify no retrieval occurs to check both successful behavior and the remaining operational boundary.

E:

Attempt prompt injection asking for another tenant's source and verify no retrieval occurs exercises a relevant negative or edge-case condition. It matters here because even isolated retrieval can expose data through shared logs or cached responses. The complementary positive check—query the same contract clause as principals from two tenants and inspect citations—then confirms the intended business capability still works; neither observation alone settles the whole requirement.

Incorrect Answers

B:

Displaying a tenant label cannot guarantee isolation of underlying chunks. For this workload, the decisive functional evidence is: Query the same contract clause as principals from two tenants and inspect citations. The shortcut in this option could pass even when a client receives a correct-looking answer citing another customer's contract.

C:

An overprivileged identity masks unauthorized retrieval that a normal tenant must not perform. The deployment needs to enforce tenant-specific grounding at the retrieval layer. An independent test is: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

D:

Post-generation filtering is too late to protect the retrieval boundary. The real incident to guard against is that a client receives a correct-looking answer citing another customer's contract. This does not replace the case-specific validation: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

Question 132

An AI engineer is documenting caveats after implementing a solution to enforce tenant-specific grounding at the retrieval layer. The workload is: A legal retrieval assistant must respond to a question about a contract without using evidence from other clients. Which residual risk deserves explicit validation?

  1. Handoff queues need service-level controls or escalations can become indefinite stalls.
  2. A second model can share systematic biases unless evidence is independently checked.
  3. Even isolated retrieval can expose data through shared logs or cached responses.
  4. Prompt compression can remove rare but essential constraints.

Explanation

Correct Answer

C:

Even after the team chooses to apply tenant identity filters in search and retain citation IDs for each retrieved passage, even isolated retrieval can expose data through shared logs or cached responses. This matters for a legal retrieval assistant must respond to a question about a contract without using evidence from other clients. The edge-case check is: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs. The regular success check alone—query the same contract clause as principals from two tenants and inspect citations—cannot establish that this risk is controlled.

Incorrect Answers

A:

Handoff queues need service-level controls or escalations can become indefinite stalls could be important if a coaching assistant must switch between automated advice and human review when topics exceed its permissions, but the actual use case is a legal retrieval assistant must respond to a question about a contract without using evidence from other clients. The closest operational risk is even isolated retrieval can expose data through shared logs or cached responses; evaluate that instead of handoff queues need service-level controls or escalations can become indefinite stalls by testing: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

B:

A second model can share systematic biases unless evidence is independently checked could be important if a high-stakes extraction model sometimes fails a confidence check and needs a second independent check, but the actual use case is a legal retrieval assistant must respond to a question about a contract without using evidence from other clients. The closest operational risk is even isolated retrieval can expose data through shared logs or cached responses; evaluate that instead of a second model can share systematic biases unless evidence is independently checked by testing: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

D:

Prompt compression can remove rare but essential constraints could be important if a conversational assistant becomes overly verbose and expensive during peak demand, but the actual use case is a legal retrieval assistant must respond to a question about a contract without using evidence from other clients. The closest operational risk is even isolated retrieval can expose data through shared logs or cached responses; evaluate that instead of prompt compression can remove rare but essential constraints by testing: Attempt prompt injection asking for another tenant's source and verify no retrieval occurs.

Question 133

A team implementing Microsoft Foundry faces this scenario: A product assistant receives free-text tool responses but needs to pass prices as validated numbers. The nonnegotiable requirement is to prevent unsafe conversion of model guesses into pricing records. Which design choice best meets it?

  1. Extract the first number anywhere in the page and treat it as price.
  2. Ask the model to choose a reasonable price when the field is missing.
  3. Skip currency conversion and silently label every amount USD.
  4. Treat tool output as untrusted, parse typed price fields, and validate currency and range before use.

Explanation

Correct Answer

D:

Tool response structure and business constraints must be checked outside generation; for this workload the decision is to treat tool output as untrusted, parse typed price fields, and validate currency and range before use. The acceptance result is: Feed mixed numeric fields and verify the structured parser selects price and currency correctly. An independent edge-case check is: Omit currency and confirm the checkout step rejects an ambiguous amount. That distinction matters because typed parsing prevents some errors but cannot verify a compromised upstream pricing service.

Incorrect Answers

A:

The first number may be a stock count or discount, which does not satisfy the requirement to prevent unsafe conversion of model guesses into pricing records. A suitable test of the correct implementation is: Feed mixed numeric fields and verify the structured parser selects price and currency correctly. This catches the particular incident where the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price.

B:

Inference cannot substitute for the authoritative price. This shortcut can leave the following observable failure unresolved: The catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. A proper deployment instead must prevent unsafe conversion of model guesses into pricing records; the missing capability would leave the deployment exposed to the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. The right approach is to treat tool output as untrusted, parse typed price fields, and validate currency and range before use, rather than relying on this unrelated shortcut. The distinguishing test is: Feed mixed numeric fields and verify the structured parser selects price and currency correctly.

C:

Currency errors can materially change the transaction, so it cannot establish prevent unsafe conversion of model guesses into pricing records. The working design instead needs to treat tool output as untrusted, parse typed price fields, and validate currency and range before use. A strong boundary test is: Omit currency and confirm the checkout step rejects an ambiguous amount. This matters because typed parsing prevents some errors but cannot verify a compromised upstream pricing service.

Question 134

After rollout, the team observes: The catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. The intended design was to prevent unsafe conversion of model guesses into pricing records. Which root cause should be investigated first?

  1. Retrieval returned chunks without enforcing their ACL or department scope.
  2. The retrieval request or index mapping lacks effective keyword matching on the code field.
  3. The parser bound a count field to the monetary amount.
  4. The application treats failed tool access as permission to infer live data.

Explanation

Correct Answer

C:

The failure that the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price points to the parser bound a count field to the monetary amount. The direct functional check is: Feed mixed numeric fields and verify the structured parser selects price and currency correctly. Then test the independent boundary—omit currency and confirm the checkout step rejects an ambiguous amount—to distinguish a genuine repair from a passing routine response.

Incorrect Answers

A:

If retrieval returned chunks without enforcing their ACL or department scope, then the expected symptom is a sales employee obtains a confidential HR paragraph in an otherwise relevant answer; this incident instead reports that the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. The nearer diagnosis is the parser bound a count field to the monetary amount; to distinguish it from the alternative of retrieval returned chunks without enforcing their ACL or department scope, test the specific service behavior: Feed mixed numeric fields and verify the structured parser selects price and currency correctly.

B:

If the retrieval request or index mapping lacks effective keyword matching on the code field, then the expected symptom is the assistant handles a symptom paraphrase but misses an exact error code mentioned in a manual; this incident instead reports that the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. The nearer diagnosis is the parser bound a count field to the monetary amount; to distinguish it from the alternative of the retrieval request or index mapping lacks effective keyword matching on the code field, test the specific service behavior: Feed mixed numeric fields and verify the structured parser selects price and currency correctly.

D:

If the application treats failed tool access as permission to infer live data, then the expected symptom is the availability API timed out and the agent continued by fabricating a delivery slot, rather than reporting that the live status was unavailable; this incident instead reports that the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. The nearer diagnosis is the parser bound a count field to the monetary amount; to distinguish it from the alternative of the application treats failed tool access as permission to infer live data, test the specific service behavior: Feed mixed numeric fields and verify the structured parser selects price and currency correctly.

Question 135

A release reviewer considers this scenario: A product assistant receives free-text tool responses but needs to pass prices as validated numbers. The objective is to prevent unsafe conversion of model guesses into pricing records. Which TWO validation outcomes directly establish that behavior? Select TWO answers.

  1. Omit currency and confirm the checkout step rejects an ambiguous amount.
  2. Check that the returned price looks plausible while ignoring whether the currency field is present.
  3. Retry a missing-currency request until a checkout record is created, rather than returning a validation error.
  4. Feed mixed numeric fields and verify the structured parser selects price and currency correctly.
  5. Test only a single isolated price field and omit mixed numeric responses containing counts and amounts.

Explanation

Correct Answers

A:

Omit currency and confirm the checkout step rejects an ambiguous amount exercises a relevant negative or edge-case condition. It matters here because typed parsing prevents some errors but cannot verify a compromised upstream pricing service. The complementary positive check—feed mixed numeric fields and verify the structured parser selects price and currency correctly—then confirms the intended business capability still works; neither observation alone settles the whole requirement.

D:

Feed mixed numeric fields and verify the structured parser selects price and currency correctly measures the actual outcome needed to prevent unsafe conversion of model guesses into pricing records. It would expose the specific failure where the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. Pair it with omit currency and confirm the checkout step rejects an ambiguous amount to check both successful behavior and the remaining operational boundary.

Incorrect Answers

B:

A plausible number cannot prove correct currency or field mapping. For this workload, the decisive functional evidence is: Feed mixed numeric fields and verify the structured parser selects price and currency correctly. The shortcut in this option could pass even when the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price.

C:

Retries do not resolve missing required parameters and may create wrong charges. The deployment needs to prevent unsafe conversion of model guesses into pricing records. An independent test is: Omit currency and confirm the checkout step rejects an ambiguous amount.

E:

The actual parser risk is binding the wrong numeric field when multiple values are present. The real incident to guard against is that the catalog service returns 19 items and 129.00 price, but the app writes 19.00 as price. This does not replace the case-specific validation: Omit currency and confirm the checkout step rejects an ambiguous amount.

Question 136

An Azure AI engineer must address the following production request: A RAG assistant must support long policies whose clauses cross page boundaries. The solution must preserve semantic context and source attribution during chunking. What should be implemented?

  1. Split by sections or overlap with stable page/section metadata and measure answer coverage.
  2. Remove heading metadata before embedding because it is not natural language.
  3. Store only the first page of each document to reduce embeddings cost.
  4. Split all PDFs into isolated single words.

Explanation

Correct Answer

A:

Meaningful chunk boundaries preserve clauses that would be lost in rigid page cuts; for this workload the decision is to split by sections or overlap with stable page/section metadata and measure answer coverage. The acceptance result is: Ask questions requiring both a clause and its adjacent exception. An independent edge-case check is: Verify citations identify both sections rather than only the leading page. That distinction matters because even semantic chunks can lose cross-document context without deliberate retrieval evaluation.

Incorrect Answers

B:

Headings provide retrieval context for similar clauses, which does not satisfy the requirement to preserve semantic context and source attribution during chunking. A suitable test of the correct implementation is: Ask questions requiring both a clause and its adjacent exception. This catches the particular incident where the retrieved chunk omits the exception beginning on the next page.

C:

Later critical clauses become unsearchable. This shortcut can leave the following observable failure unresolved: The retrieved chunk omits the exception beginning on the next page. A proper deployment instead must preserve semantic context and source attribution during chunking; the missing capability would leave the deployment exposed to the retrieved chunk omits the exception beginning on the next page. The right approach is to split by sections or overlap with stable page/section metadata and measure answer coverage, rather than relying on this unrelated shortcut. The distinguishing test is: Ask questions requiring both a clause and its adjacent exception.

D:

Tiny fragments destroy the relation among conditions, so it cannot establish preserve semantic context and source attribution during chunking. The working design instead needs to split by sections or overlap with stable page/section metadata and measure answer coverage. A strong boundary test is: Verify citations identify both sections rather than only the leading page. This matters because even semantic chunks can lose cross-document context without deliberate retrieval evaluation.

Question 137

A Foundry application has a specific incident: The retrieved chunk omits the exception beginning on the next page. The operator is trying to preserve semantic context and source attribution during chunking. Which failure mechanism best explains the evidence?

  1. Retrieval returned chunks without enforcing their ACL or department scope.
  2. Application validation lacks the enumerated allowed values or schema checks.
  3. Chunk boundaries severed a clause from its limiting condition.
  4. The retrieval request or index mapping lacks effective keyword matching on the code field.

Explanation

Correct Answer

C:

The failure that the retrieved chunk omits the exception beginning on the next page points to chunk boundaries severed a clause from its limiting condition. The direct functional check is: Ask questions requiring both a clause and its adjacent exception. Then test the independent boundary—verify citations identify both sections rather than only the leading page—to distinguish a genuine repair from a passing routine response.

Incorrect Answers

A:

If retrieval returned chunks without enforcing their ACL or department scope, then the expected symptom is a sales employee obtains a confidential HR paragraph in an otherwise relevant answer; this incident instead reports that the retrieved chunk omits the exception beginning on the next page. The nearer diagnosis is chunk boundaries severed a clause from its limiting condition; to distinguish it from the alternative of retrieval returned chunks without enforcing their ACL or department scope, test the specific service behavior: Ask questions requiring both a clause and its adjacent exception.

B:

If application validation lacks the enumerated allowed values or schema checks, then the expected symptom is the model returns valid JSON with a category not recognized by downstream routing; this incident instead reports that the retrieved chunk omits the exception beginning on the next page. The nearer diagnosis is chunk boundaries severed a clause from its limiting condition; to distinguish it from the alternative of application validation lacks the enumerated allowed values or schema checks, test the specific service behavior: Ask questions requiring both a clause and its adjacent exception.

D:

If the retrieval request or index mapping lacks effective keyword matching on the code field, then the expected symptom is the assistant handles a symptom paraphrase but misses an exact error code mentioned in a manual; this incident instead reports that the retrieved chunk omits the exception beginning on the next page. The nearer diagnosis is chunk boundaries severed a clause from its limiting condition; to distinguish it from the alternative of the retrieval request or index mapping lacks effective keyword matching on the code field, test the specific service behavior: Ask questions requiring both a clause and its adjacent exception.

Question 138

A reviewer must verify that the solution will preserve semantic context and source attribution during chunking. The deployment context is: A RAG assistant must support long policies whose clauses cross page boundaries. Which TWO tests provide the strongest evidence? Select TWO answers.

  1. Verify that the summary includes the clause title but do not include its limiting exception.
  2. Verify citations identify both sections rather than only the leading page.
  3. Check that two passages rank highly in retrieval but never compare the generated claim against either passage.
  4. Provide a citation to the first page only when the exception starts on the next page.
  5. Ask questions requiring both a clause and its adjacent exception.

Explanation

Correct Answers

B:

Verify citations identify both sections rather than only the leading page exercises a relevant negative or edge-case condition. It matters here because even semantic chunks can lose cross-document context without deliberate retrieval evaluation. The complementary positive check—ask questions requiring both a clause and its adjacent exception—then confirms the intended business capability still works; neither observation alone settles the whole requirement.

E:

Ask questions requiring both a clause and its adjacent exception measures the actual outcome needed to preserve semantic context and source attribution during chunking. It would expose the specific failure where the retrieved chunk omits the exception beginning on the next page. Pair it with verify citations identify both sections rather than only the leading page to check both successful behavior and the remaining operational boundary.

Incorrect Answers

A:

A correct heading can still materially misstate a contract condition. For this workload, the decisive functional evidence is: Ask questions requiring both a clause and its adjacent exception. The shortcut in this option could pass even when the retrieved chunk omits the exception beginning on the next page.

C:

Retrieval ranking does not prove the answer accounts for the exception. The deployment needs to preserve semantic context and source attribution during chunking. An independent test is: Verify citations identify both sections rather than only the leading page.

D:

A citation missing the exception cannot support the whole interpreted clause. The real incident to guard against is that the retrieved chunk omits the exception beginning on the next page. This does not replace the case-specific validation: Verify citations identify both sections rather than only the leading page.

Question 139

An AI engineer is documenting caveats after implementing a solution to preserve semantic context and source attribution during chunking. The workload is: A RAG assistant must support long policies whose clauses cross page boundaries. Which residual risk deserves explicit validation?

  1. Routing errors can cost more than savings if failures require repeated calls.
  2. Distributed trace gaps and sampling can hide rare tail latency events.
  3. A second model can share systematic biases unless evidence is independently checked.
  4. Even semantic chunks can lose cross-document context without deliberate retrieval evaluation.

Explanation

Correct Answer

D:

Even after the team chooses to split by sections or overlap with stable page/section metadata and measure answer coverage, even semantic chunks can lose cross-document context without deliberate retrieval evaluation. This matters for a RAG assistant must support long policies whose clauses cross page boundaries. The edge-case check is: Verify citations identify both sections rather than only the leading page. The regular success check alone—ask questions requiring both a clause and its adjacent exception—cannot establish that this risk is controlled.

Incorrect Answers

A:

Routing errors can cost more than savings if failures require repeated calls could be important if a multilingual assistant uses both a small classifier and a larger generative model, but the actual use case is a RAG assistant must support long policies whose clauses cross page boundaries. The closest operational risk is even semantic chunks can lose cross-document context without deliberate retrieval evaluation; evaluate that instead of routing errors can cost more than savings if failures require repeated calls by testing: Verify citations identify both sections rather than only the leading page.

B:

Distributed trace gaps and sampling can hide rare tail latency events could be important if a customer-service agent shows a two-second model response time but an eight-second end-to-end delay, but the actual use case is a RAG assistant must support long policies whose clauses cross page boundaries. The closest operational risk is even semantic chunks can lose cross-document context without deliberate retrieval evaluation; evaluate that instead of distributed trace gaps and sampling can hide rare tail latency events by testing: Verify citations identify both sections rather than only the leading page.

C:

A second model can share systematic biases unless evidence is independently checked could be important if a high-stakes extraction model sometimes fails a confidence check and needs a second independent check, but the actual use case is a RAG assistant must support long policies whose clauses cross page boundaries. The closest operational risk is even semantic chunks can lose cross-document context without deliberate retrieval evaluation; evaluate that instead of a second model can share systematic biases unless evidence is independently checked by testing: Verify citations identify both sections rather than only the leading page.

Question 140

Consider an application deployed with Microsoft Foundry: A finance extraction app must cite sources and reject answers whose values are not present in retrieved statements. The owner wants to control hallucinations with evidence-based acceptance criteria. Which implementation is most appropriate?

  1. Treat a fluent answer as correct if it uses accounting terminology.
  2. Disable retrieval for documents older than one day regardless of requested reporting period.
  3. Ask the model to supply the most likely missing balance.
  4. Compare generated financial fields to retrieved statement spans and abstain when a required value is absent.

Explanation

Correct Answer

D:

Absent evidence should yield an explicit unknown, not an invented number; for this workload the decision is to compare generated financial fields to retrieved statement spans and abstain when a required value is absent. The acceptance result is: Remove the balance span and verify the answer reports insufficient evidence. An independent edge-case check is: Restore the span and verify the returned number matches exact cited units. That distinction matters because an extracted number can be correct while the reporting period or unit is misattributed.

Incorrect Answers

A:

Terminology does not validate numbers, which does not satisfy the requirement to control hallucinations with evidence-based acceptance criteria. A suitable test of the correct implementation is: Remove the balance span and verify the answer reports insufficient evidence. This catches the particular incident where a monthly balance is shown although no ledger document contains that figure.

B:

Historical statements may be the authoritative period. This shortcut can leave the following observable failure unresolved: A monthly balance is shown although no ledger document contains that figure. A proper deployment instead must control hallucinations with evidence-based acceptance criteria; the missing capability would leave the deployment exposed to a monthly balance is shown although no ledger document contains that figure. The right approach is to compare generated financial fields to retrieved statement spans and abstain when a required value is absent, rather than relying on this unrelated shortcut. The distinguishing test is: Remove the balance span and verify the answer reports insufficient evidence.

C:

Estimated values are not source facts, so it cannot establish control hallucinations with evidence-based acceptance criteria. The working design instead needs to compare generated financial fields to retrieved statement spans and abstain when a required value is absent. A strong boundary test is: Restore the span and verify the returned number matches exact cited units. This matters because an extracted number can be correct while the reporting period or unit is misattributed.

Developing AI Apps and Agents on Azure (AI-103) Practice Tests

SPECIAL OFFER: GET 10% OFF
This is ONE TIME OFFER

You save
10%

Enter Your Email Address to Receive Your 10% Off Discount Code

SPECIAL OFFER: GET 10% OFF

You save
10%

Use Discount Code:

A confirmation link was sent to your e-mail.

Please check your mailbox for a message from support@examlabs.com and follow the directions.

Download Free Demo of VCE Exam Simulator

Experience Avanset VCE Exam Simulator for yourself.

Simply submit your email address below to get started with our interactive software demo of your free trial.

  • Realistic exam simulation and exam editor with preview functions
  • Whole exam in a single file with several different question types
  • Customizable exam-taking mode & detailed score reports