cert
cert-1
cert-2

Pass Microsoft AI-300 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
AI-300 Exam - Verified By Experts
AI-300 Premium File

AI-300 Premium File

$69.99
$76.99
  • Premium File 213 Questions & Answers. Last Update: Nov 12, 2026

Whats Included:

  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
 
$76.99
$69.99
accept 45 downloads in the last 7 days
block-screenshots
Microsoft AI-300 Practice Test 57

Question 560

During a production Azure Machine Learning model lifecycle incident, the following condition occurs: An engineer promotes the model with the lowest training loss even though the held-out validation loss worsens, indicating overfitting. The operator must safeguard training reproducibility and serving continuity. Which diagnostic evidence most directly distinguishes the described AI operations failure?

  1. Check mlflow.autolog arguments, run artifact listing, model flavor metadata, configured output path and registration job logs.
  2. Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics.
  3. Review distributed sampler configuration, global and local ranks, dataset partition checks, epoch seeding and examples processed per worker.
  4. Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts.

Explanation

Correct Answer

B:

Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics. This is the most relevant evidence for new candidate chosen from training instead of validation results. A training objective measures fit on observed examples; generalization claims require independent validation and selection rules. The matching controlled remediation is: Use a documented holdout or cross-validation-based model selection procedure and require a production-relevant promotion threshold.

Incorrect Answers

A:

Check mlflow.autolog arguments, run artifact listing, model flavor metadata, configured output path and registration job logs. This would investigate autologging excludes required model artifact rather than the documented new candidate chosen from training instead of validation results issue. The relevant investigation is: Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics.

C:

Review distributed sampler configuration, global and local ranks, dataset partition checks, epoch seeding and examples processed per worker. This would investigate distributed training uses one shard for every worker rather than the documented new candidate chosen from training instead of validation results issue. The relevant investigation is: Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics.

D:

Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. This would investigate endpoint latency average conceals tail degradation rather than the documented new candidate chosen from training instead of validation results issue. The relevant investigation is: Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics.

Question 561

As an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A data scientist runs a local notebook intending to log results to a governed Azure ML workspace, but the MLflow tracking URI still points to an earlier sandbox workspace. The resolution must preserve quality gates, documented artifacts and rollback options. What should the engineer examine first to confirm the specific AI-300 workload issue?

  1. Inspect online endpoint traffic weights, deployment versions, request tracing, rollout approval and available prior stable deployment.
  2. Inspect distributed job settings, node count, rendezvous endpoint configuration, compute network security rules and worker log timeouts.
  3. Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability.
  4. Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process.

Explanation

Correct Answer

D:

Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process. This is the most relevant evidence for mlflow tracking uri sends notebook runs to wrong workspace. An interactive MLflow process can log to whichever tracking server is configured; Azure workspace selection is not inferred from the notebook file's location. The matching controlled remediation is: Set the correct Azure Machine Learning MLflow tracking URI and experiment context before training and verify authenticated workspace access.

Incorrect Answers

A:

Inspect online endpoint traffic weights, deployment versions, request tracing, rollout approval and available prior stable deployment. This would investigate canary rollout sends all traffic to unvalidated candidate rather than the documented mlflow tracking uri sends notebook runs to wrong workspace issue. The relevant investigation is: Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process.

B:

Inspect distributed job settings, node count, rendezvous endpoint configuration, compute network security rules and worker log timeouts. This would investigate distributed job workers cannot communicate over required network path rather than the documented mlflow tracking uri sends notebook runs to wrong workspace issue. The relevant investigation is: Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process.

C:

Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability. This would investigate automated ml experiment uses leakage feature rather than the documented mlflow tracking uri sends notebook runs to wrong workspace issue. The relevant investigation is: Inspect active MLflow tracking URI, experiment ID, Azure credentials, studio run listing and configuration inherited by the notebook process.

Question 562

During a regulated Azure ML workspace infrastructure review, the following condition occurs: A pipeline component is published for reuse but includes a hard-coded workspace datastore URI, causing executions in a second workspace to read an unavailable path. The team must preserve approved access, model provenance and private network constraints. Which targeted Azure AI operations change best addresses the documented problem?

  1. Pin the pipeline input to an approved versioned data asset and advance versions through controlled evaluation gates.
  2. Add Bicep validation, targeted what-if reviews and a controlled environment promotion stage before production apply.
  3. Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline.
  4. Grant controlled registry consumption rights and use the fully qualified versioned registry asset reference in the target workspace.

Explanation

Correct Answer

C:

Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline. This targets reusable component hardcodes workspace-specific asset path. Portable Azure ML components describe computation and typed interfaces rather than assuming a single workspace's storage layout. Verify the failure through: Inspect component YAML for workspace IDs, hard-coded URIs, parameter definitions and actual target workspace data asset registrations. Validate success with: The same versioned component runs against authorized data in two workspaces without editing its source implementation.

Incorrect Answers

A:

Pin the pipeline input to an approved versioned data asset and advance versions through controlled evaluation gates. This could address versionless dataset reference breaks reproducibility, but it does not solve the specified reusable component hardcodes workspace-specific asset path failure. The matching remediation is: Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline.

B:

Add Bicep validation, targeted what-if reviews and a controlled environment promotion stage before production apply. This could address iac workflow lacks repeatable infrastructure validation, but it does not solve the specified reusable component hardcodes workspace-specific asset path failure. The matching remediation is: Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline.

D:

Grant controlled registry consumption rights and use the fully qualified versioned registry asset reference in the target workspace. This could address registry model asset not available to second workspace, but it does not solve the specified reusable component hardcodes workspace-specific asset path failure. The matching remediation is: Replace embedded workspace-specific paths with parameterized, typed data inputs and bind them explicitly in each consuming pipeline.

Question 563

As an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: A real-time predictor appears healthy by mean request latency while high-value users face multi-second waits at the 99th percentile after a traffic spike. The resolution must preserve quality gates, documented artifacts and rollback options. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Review monitoring window, input sample count, drift statistic, alert routing and retraining schedule history.
  2. Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts.
  3. Package and version the approved feature retrieval specification with the model and enforce the same mapping at serving time.
  4. Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck.
  5. Monitor available proxy signals immediately and compute labeled performance only after verified outcomes arrive through a governed join.

Explanation

Correct Answers

B:

Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. This is the most relevant evidence for endpoint latency average conceals tail degradation. Average endpoint latency can hide slow outliers; tail distributions and cohort behavior are needed for user-facing service assurance. The matching controlled remediation is: Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck.

D:

Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck. This targets endpoint latency average conceals tail degradation. Average endpoint latency can hide slow outliers; tail distributions and cohort behavior are needed for user-facing service assurance. Verify the failure through: Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. Validate success with: P95/P99 and error-rate service objectives recover under representative peak-load traffic without unrelated scale changes.

Incorrect Answers

A:

Review monitoring window, input sample count, drift statistic, alert routing and retraining schedule history. This is associated with retraining alarm triggers on insufficient sample volume, not the essential evidence-and-fix pair for endpoint latency average conceals tail degradation. First examine: Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. Then correct: Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck.

C:

Package and version the approved feature retrieval specification with the model and enforce the same mapping at serving time. This is associated with feature retrieval specification missing from registered model, not the essential evidence-and-fix pair for endpoint latency average conceals tail degradation. First examine: Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. Then correct: Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck.

E:

Monitor available proxy signals immediately and compute labeled performance only after verified outcomes arrive through a governed join. This is associated with accuracy monitoring lacks delayed outcome labels, not the essential evidence-and-fix pair for endpoint latency average conceals tail degradation. First examine: Examine Azure Monitor request latency percentiles, response codes, deployment load, compute utilization and per-client cohorts. Then correct: Configure tail-latency and error-rate dashboards with capacity remedies appropriate to the measured bottleneck.

Question 564

While preparing a secure Foundry model or agent release, the following condition occurs: A global customer-support agent project is provisioned in a region that does not offer the selected foundation model deployment type, preventing its evaluation pipeline from starting. The remediation must meet enterprise access and reliability requirements without a broad bypass. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes.
  2. Permit only the required documented tool and service destinations via supported private or managed-network egress controls.
  3. Deploy the governed Foundry resource and project in a supported compliant region with an available model option.
  4. Select an approved multimodal model or add a supported image-to-text preprocessing service before the text generation step.
  5. Audit Foundry network ACLs, private endpoint connections, VNet peerings, firewall rules and effective caller identities.

Explanation

Correct Answers

A:

Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes. This is the most relevant evidence for foundry project created in mismatched resource location. A Foundry project depends on its parent resource location and model/service availability; creating a project does not guarantee every model is deployable there. The matching controlled remediation is: Deploy the governed Foundry resource and project in a supported compliant region with an available model option. This matters here because while preparing a secure Foundry model or agent release, the following condition occurs: A global customer-support….

C:

Deploy the governed Foundry resource and project in a supported compliant region with an available model option. This targets foundry project created in mismatched resource location. A Foundry project depends on its parent resource location and model/service availability; creating a project does not guarantee every model is deployable there. Verify the failure through: Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes. Validate success with: The project invokes the approved model endpoint and meets required data-residency policy without an unapproved region fallback. This matters here because while preparing a secure Foundry model or agent release, the following condition occurs: A global customer-support….

Incorrect Answers

B:

Permit only the required documented tool and service destinations via supported private or managed-network egress controls. This is associated with agent tool calls fail behind overbroad outbound block, not the essential evidence-and-fix pair for foundry project created in mismatched resource location. First examine: Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes. Then correct: Deploy the governed Foundry resource and project in a supported compliant region with an available model option.

D:

Select an approved multimodal model or add a supported image-to-text preprocessing service before the text generation step. This is associated with vision input sent to text-only foundation model, not the essential evidence-and-fix pair for foundry project created in mismatched resource location. First examine: Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes. Then correct: Deploy the governed Foundry resource and project in a supported compliant region with an available model option.

E:

Audit Foundry network ACLs, private endpoint connections, VNet peerings, firewall rules and effective caller identities. This is associated with network allowlist exposes production model endpoint to test subnet, not the essential evidence-and-fix pair for foundry project created in mismatched resource location. First examine: Inspect Foundry project and resource regions, model deployment regional availability, enterprise residency constraints and API error codes. Then correct: Deploy the governed Foundry resource and project in a supported compliant region with an available model option.

Question 565

During a regulated Azure ML workspace infrastructure review, the following condition occurs: A reusable feature-engineering component emits a URI folder while a downstream training step declares a tabular MLTable input, causing type validation to fail. The team must preserve approved access, model provenance and private network constraints. Which acceptance test shows the AI operations defect is corrected?

  1. The job runs with a traceable environment digest and no surprise dependency upgrade on repeated execution.
  2. Repeated deployments converge on a valid workspace and all linked dependencies without manual reruns or temporary public access.
  3. Replaying a prior run uses identical input version and produces traceable comparison metrics before a new snapshot is promoted.
  4. The pipeline resolves component dependencies and the training step reads the expected typed asset without implicit file guessing.

Explanation

Correct Answer

D:

The pipeline resolves component dependencies and the training step reads the expected typed asset without implicit file guessing. Azure ML pipeline components have explicit typed ports; connecting incompatible artifact interfaces does not automatically transform the data. a direct corrective action is to align the component output with the declared downstream input type or introduce a supported data conversion step with versioned schema.

Incorrect Answers

A:

The job runs with a traceable environment digest and no surprise dependency upgrade on repeated execution. That addresses a different configuration issue rather than the main requirement. Verify instead: The pipeline resolves component dependencies and the training step reads the expected typed asset without implicit file guessing.

B:

Repeated deployments converge on a valid workspace and all linked dependencies without manual reruns or temporary public access. That addresses a different configuration issue rather than the main requirement. Verify instead: The pipeline resolves component dependencies and the training step reads the expected typed asset without implicit file guessing.

C:

Replaying a prior run uses identical input version and produces traceable comparison metrics before a new snapshot is promoted. That addresses a different configuration issue rather than the main requirement. Verify instead: The pipeline resolves component dependencies and the training step reads the expected typed asset without implicit file guessing.

Question 566

As an AI operations team checks production agent evaluation evidence, the following condition occurs: A Foundry quality monitor aggregates responses from old and new agent deployments, masking a groundedness regression in the new version. The workflow must produce measurable quality, safety and performance evidence. Which TWO complementary choices identify the issue and correct it? Select TWO.

  1. Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples.
  2. Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results.
  3. Inspect tokens per request, input/output token ratios, concurrency, throttling status and model deployment quota.
  4. Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content.
  5. Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.

Explanation

Correct Answers

B:

Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. This is the most relevant evidence for foundry continuous evaluation ignores production deployment version. Aggregate monitoring can hide release-specific failures; quality assessment needs traceable version attribution. The matching controlled remediation is: Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.

E:

Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate. This targets foundry continuous evaluation ignores production deployment version. Aggregate monitoring can hide release-specific failures; quality assessment needs traceable version attribution. Verify the failure through: Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. Validate success with: A canary regression is visible independently of stable baseline traffic before expanding rollout.

Incorrect Answers

A:

Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples. This is associated with prompt truncation strategy increases generated-token spend, not the essential evidence-and-fix pair for foundry continuous evaluation ignores production deployment version. First examine: Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. Then correct: Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.

C:

Inspect tokens per request, input/output token ratios, concurrency, throttling status and model deployment quota. This is associated with genai throughput report mixes responses and tokens, not the essential evidence-and-fix pair for foundry continuous evaluation ignores production deployment version. First examine: Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. Then correct: Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.

D:

Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content. This is associated with harmful content probe set excludes multilingual input, not the essential evidence-and-fix pair for foundry continuous evaluation ignores production deployment version. First examine: Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. Then correct: Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.

Question 567

During a regulated Azure ML workspace infrastructure review, the following condition occurs: A new Azure ML workspace starts but managed training cannot retrieve approved secrets because deployment references a Key Vault belonging to an unrelated project. The team must preserve approved access, model provenance and private network constraints. Which observable postchange result proves the intended remediation succeeded?

  1. A scoped job retrieves only the approved secret and unrelated project secrets remain inaccessible.
  2. Authorized workspaces run the shared component while unauthorized subscriptions remain blocked from the registry.
  3. The job consumes the intended dataset version and staging files are not silently mixed into production runs.
  4. The production team creates and reads workspace assets under the correct policy scope while sandbox users cannot administer production resources.

Explanation

Correct Answer

A:

A scoped job retrieves only the approved secret and unrelated project secrets remain inaccessible. A workspace can exist while supporting Key Vault, storage or monitoring dependencies are incorrectly associated; successful resource creation alone is not end-to-end readiness. a direct corrective action is to link the authorized Key Vault or repair supported access to the intended secret store through the correct workspace deployment configuration.

Incorrect Answers

B:

Authorized workspaces run the shared component while unauthorized subscriptions remain blocked from the registry. That addresses a different configuration issue rather than the main requirement. Verify instead: A scoped job retrieves only the approved secret and unrelated project secrets remain inaccessible.

C:

The job consumes the intended dataset version and staging files are not silently mixed into production runs. That addresses a different configuration issue rather than the main requirement. Verify instead: A scoped job retrieves only the approved secret and unrelated project secrets remain inaccessible.

D:

The production team creates and reads workspace assets under the correct policy scope while sandbox users cannot administer production resources. That addresses a different configuration issue rather than the main requirement. Verify instead: A scoped job retrieves only the approved secret and unrelated project secrets remain inaccessible.

Question 568

While deploying reproducible Azure Machine Learning infrastructure, the following condition occurs: A GitHub federated identity credential accepts tokens from any branch, allowing an unreviewed pull request to reach Azure Machine Learning production resources. The change must remain limited to authorized workspaces and avoid unreviewed production impact. What approved correction resolves the failure without modifying unrelated healthy services?

  1. Add Bicep validation, targeted what-if reviews and a controlled environment promotion stage before production apply.
  2. Require protected pull requests and automated component tests before changes to production training code are merged.
  3. Restrict the federated identity subject to the protected deployment environment or branch and enforce repository approval rules.
  4. Scope an approved least-privilege Azure ML role to the intended resource group or workspace and remove the broad inherited Contributor grant.

Explanation

Correct Answer

C:

Restrict the federated identity subject to the protected deployment environment or branch and enforce repository approval rules. This targets oidc subject filter trusts unreviewed pull-request branches. Federated authentication remains vulnerable to overbroad trust claims if token subject matching does not reflect deployment governance. Verify the failure through: Review federated credential issuer, audience and subject, GitHub environment protection rules, branch patterns and run provenance. Validate success with: Only authorized protected workflows obtain production Azure tokens while unreviewed pull requests fail the access test. This matters here because while deploying reproducible Azure Machine Learning infrastructure, the following condition occurs: A GitHub….

Incorrect Answers

A:

Add Bicep validation, targeted what-if reviews and a controlled environment promotion stage before production apply. This could address iac workflow lacks repeatable infrastructure validation, but it does not solve the specified oidc subject filter trusts unreviewed pull-request branches failure. The matching remediation is: Restrict the federated identity subject to the protected deployment environment or branch and enforce repository approval rules. This matters here because while deploying reproducible Azure Machine Learning infrastructure, the following condition occurs: A GitHub….

B:

Require protected pull requests and automated component tests before changes to production training code are merged. This could address notebook merge bypasses protected evaluation tests, but it does not solve the specified oidc subject filter trusts unreviewed pull-request branches failure. The matching remediation is: Restrict the federated identity subject to the protected deployment environment or branch and enforce repository approval rules.

D:

Scope an approved least-privilege Azure ML role to the intended resource group or workspace and remove the broad inherited Contributor grant. This could address workspace contributor role granted to untrusted automation, but it does not solve the specified oidc subject filter trusts unreviewed pull-request branches failure. The matching remediation is: Restrict the federated identity subject to the protected deployment environment or branch and enforce repository approval rules.

Question 569

While optimizing a production RAG or fine-tuned model service, the following condition occurs: A CI pipeline immediately swaps a production fine-tuned model for each completed experiment without validating cost, quality and deployment compatibility. The resulting model must meet held-out quality and security criteria. What approved correction resolves the failure without modifying unrelated healthy services?

  1. Rebalance curated synthetic and real cases to preserve rare operational scenarios and evaluate on an untouched real-world edge-case set.
  2. Expand representative domain pairs and negatives, apply appropriate embedding model adaptation and retain a separate general-language retrieval evaluation.
  3. Generate sanitized synthetic patterns from approved nonidentifying templates and screen outputs for memorized or copied private content.
  4. Register fine-tuned candidates as versioned artifacts, enforce evaluation and security gates, then canary only approved versions.

Explanation

Correct Answer

D:

Register fine-tuned candidates as versioned artifacts, enforce evaluation and security gates, then canary only approved versions. This targets fine-tuned candidate bypasses promotion review after training. Training completion is a research milestone, not authorization for production; fine-tuned releases need controlled model registration, approval and rollout. Verify the failure through: Examine fine-tuning run outputs, model registry versions, promotion approvals, held-out evaluations and rollout routing settings. Validate success with: A bad candidate remains isolated while an approved model reaches production with known-good rollback and auditable comparison results. This matters here because while optimizing a production RAG or fine-tuned model service, the following condition occurs: A CI pipeline….

Incorrect Answers

A:

Rebalance curated synthetic and real cases to preserve rare operational scenarios and evaluate on an untouched real-world edge-case set. This could address synthetic data overwhelms minority real-world edge cases, but it does not solve the specified fine-tuned candidate bypasses promotion review after training failure. The matching remediation is: Register fine-tuned candidates as versioned artifacts, enforce evaluation and security gates, then canary only approved versions.

B:

Expand representative domain pairs and negatives, apply appropriate embedding model adaptation and retain a separate general-language retrieval evaluation. This could address specialized embedding adaptation overfits small terminology set, but it does not solve the specified fine-tuned candidate bypasses promotion review after training failure. The matching remediation is: Register fine-tuned candidates as versioned artifacts, enforce evaluation and security gates, then canary only approved versions.

C:

Generate sanitized synthetic patterns from approved nonidentifying templates and screen outputs for memorized or copied private content. This could address synthetic training examples copy private customer identifiers, but it does not solve the specified fine-tuned candidate bypasses promotion review after training failure. The matching remediation is: Register fine-tuned candidates as versioned artifacts, enforce evaluation and security gates, then canary only approved versions. This matters here because while optimizing a production RAG or fine-tuned model service, the following condition occurs: A CI pipeline….

Question 570

During a Foundry generative AI quality and observability incident, the following condition occurs: An enterprise support agent appears successful on isolated questions but fails follow-up conversations because the test dataset discards prior chat turns and tool context. The fix must preserve evaluation validity and avoid unnecessary user-data collection. Which investigation best verifies the cause before changing production configuration?

  1. Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.
  2. Inspect safety evaluator coverage, adversarial retrieval examples, trusted instruction hierarchy, tool-action traces and current release criteria.
  3. Examine supported language list, safety evaluation examples, category coverage, severity thresholds and live red-team evidence.
  4. Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results.

Explanation

Correct Answer

A:

Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations. This is the most relevant evidence for multi-turn agent evaluation uses isolated last message only. Multi-turn agents depend on conversation state; evaluating only the final prompt can miss context handling, memory and tool-use errors. The matching controlled remediation is: Include appropriate complete conversation histories and tool events in the supported agent evaluation workflow while preserving privacy protections.

Incorrect Answers

B:

Inspect safety evaluator coverage, adversarial retrieval examples, trusted instruction hierarchy, tool-action traces and current release criteria. This would investigate content safety evaluation omits prompt injection tests rather than the documented multi-turn agent evaluation uses isolated last message only issue. The relevant investigation is: Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.

C:

Examine supported language list, safety evaluation examples, category coverage, severity thresholds and live red-team evidence. This would investigate harmful content probe set excludes multilingual input rather than the documented multi-turn agent evaluation uses isolated last message only issue. The relevant investigation is: Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.

D:

Examine continuous monitoring configuration, trace deployment IDs, sampling window, evaluator assignment and per-version quality results. This would investigate foundry continuous evaluation ignores production deployment version rather than the documented multi-turn agent evaluation uses isolated last message only issue. The relevant investigation is: Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.

Operationalizing Machine Learning and Generative AI Solutions (AI-300) Practice Tests