Question 1040A customer churn model sees a new categorical distribution after the company launches in another country. Which measurement or test should the team perform after implementation? Apply a safe controlled load and verify p95 latency, error-rate and resource alarms trigger with clear ownership. Compare sampled source citations and factuality against approved references before and after the index update. Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected. Compare local attribution stability on similar cases and check against controlled perturbations and business rules. ExplanationCorrect Answer C:Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected. This tests the underlying failure: Data drift is a change in input distributions and can occur even when prediction-outcome labels are delayed. Other Answers A:Reliability metrics expose serving failures independently from predictive model quality and training drift. That does not establish the required outcome; verify it by checking: Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected. B:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. That does not establish the required outcome; verify it by checking: Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected. D:Feature attribution describes how inputs relate to predictions, but explanations are not necessarily causal proof. That does not establish the required outcome; verify it by checking: Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected.
Question 1041A customer churn model sees a new categorical distribution after the company launches in another country. Which ongoing standard would make this result reliable for future model versions? Version retrieval indexes and evaluation prompts and gate releases on grounded-answer regression tests. Refresh reference distributions only after approved investigation and maintain source-drift incident runbooks. Require fairness monitoring, documented mitigation ownership and human escalation for high-impact decisions. Maintain separate service reliability and model-quality SLOs with tested incident escalation paths. ExplanationCorrect Answer B:Refresh reference distributions only after approved investigation and maintain source-drift incident runbooks. This protects the selected approach because Data drift is a change in input distributions and can occur even when prediction-outcome labels are delayed. The immediate solution is: Configure Agent Platform Model Monitoring to alert on meaningful feature-distribution drift from a valid reference dataset. Other Answers A:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. The continuing safeguard required here is: Refresh reference distributions only after approved investigation and maintain source-drift incident runbooks. C:Aggregate accuracy can obscure uneven subgroup errors and biased training labels or sampling distributions. The continuing safeguard required here is: Refresh reference distributions only after approved investigation and maintain source-drift incident runbooks. D:Reliability metrics expose serving failures independently from predictive model quality and training drift. The continuing safeguard required here is: Refresh reference distributions only after approved investigation and maintain source-drift incident runbooks.
Question 1042A classifier's customer outcomes change even though its input feature distribution appears stable. Which proposed approach best resolves the described model engineering problem? Restrict the workload service account to least-privilege roles and the minimum approved resource set. Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. Run continuous groundedness, citation relevance and task-completion evaluations on a held-out generative AI dataset. Instrument request latency, availability, utilization and queueing with production Model Monitoring and cloud telemetry. ExplanationCorrect Answer B:Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. To validate the correction: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. Other Answers A:IAM controls cloud API permissions independently of whether a model produces accurate or safe predictions. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. C:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. D:Reliability metrics expose serving failures independently from predictive model quality and training drift. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone.
Question 1043A classifier's customer outcomes change even though its input feature distribution appears stable. Which property of the selected ML workflow accounts for this requirement? Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. Model instructions are probabilistic and safety guardrails need separate policy checks at input and output boundaries. Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. Reliability metrics expose serving failures independently from predictive model quality and training drift. ExplanationCorrect Answer A:Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. This supports the specific implementation: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. A suitable independent check is: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. Other Answers B:Model instructions are probabilistic and safety guardrails need separate policy checks at input and output boundaries. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. C:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. D:Reliability metrics expose serving failures independently from predictive model quality and training drift. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals.
Question 1044A classifier's customer outcomes change even though its input feature distribution appears stable. What is the most relevant acceptance check for the proposed ML change? Apply a safe controlled load and verify p95 latency, error-rate and resource alarms trigger with clear ownership. Replay shifted fixture data and confirm configured drift thresholds and alert routing work as expected. Run allowed and denied access tests using the actual serving identity and audit role grants. Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. ExplanationCorrect Answer D:Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. This tests the underlying failure: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. Other Answers A:Reliability metrics expose serving failures independently from predictive model quality and training drift. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. B:Data drift is a change in input distributions and can occur even when prediction-outcome labels are delayed. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. C:IAM controls cloud API permissions independently of whether a model produces accurate or safe predictions. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules.
Question 1045A classifier's customer outcomes change even though its input feature distribution appears stable. What preventive release checklist item best protects the intended outcome? Require preprocessing digest parity and contract tests for every model and feature pipeline update. Apply data-classification and logging reviews before enabling verbose AI application diagnostics. Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. Define attribution-change alert policies and require domain analysis before model replacement. ExplanationCorrect Answer C:Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. This protects the selected approach because Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. The immediate solution is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. Other Answers A:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. B:Observability data can itself expose confidential training or inference content if logs are collected without safeguards. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. D:Attribution drift can signal changes in how a model uses information that overall aggregate metrics may miss. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation.
Question 1046A fraud prediction model's true error rates rise because attackers adapt their behavior to old decision rules. Which implementation response is most appropriate for this ML engineering situation? Instrument request latency, availability, utilization and queueing with production Model Monitoring and cloud telemetry. Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. Restrict the workload service account to least-privilege roles and the minimum approved resource set. Add training-serving feature-skew checks and deploy the exact versioned preprocessing artifacts used during training. ExplanationCorrect Answer B:Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. To validate the correction: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. Other Answers A:Reliability metrics expose serving failures independently from predictive model quality and training drift. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. C:IAM controls cloud API permissions independently of whether a model produces accurate or safe predictions. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. D:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone.
Question 1047A fraud prediction model's true error rates rise because attackers adapt their behavior to old decision rules. Which underlying ML principle makes the proposed approach effective? Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. Feature attribution describes how inputs relate to predictions, but explanations are not necessarily causal proof. Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. Training-serving skew can degrade predictions without any change to the underlying fitted model weights. ExplanationCorrect Answer C:Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. This supports the specific implementation: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. A suitable independent check is: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. Other Answers A:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. B:Feature attribution describes how inputs relate to predictions, but explanations are not necessarily causal proof. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. D:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. The cause in the stated scenario is different: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals.
Question 1048A fraud prediction model's true error rates rise because attackers adapt their behavior to old decision rules. What is the most relevant acceptance check for the proposed ML change? Replay identical raw cases in both environments and compare feature vectors and model outputs. Compare group metrics and representative false negatives on consented and legally appropriate audit samples. Compare sampled source citations and factuality against approved references before and after the index update. Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. ExplanationCorrect Answer D:Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. This tests the underlying failure: Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. Other Answers A:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. B:Aggregate accuracy can obscure uneven subgroup errors and biased training labels or sampling distributions. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. C:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. That does not establish the required outcome; verify it by checking: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules.
Question 1049A fraud prediction model's true error rates rise because attackers adapt their behavior to old decision rules. Which ongoing standard would make this result reliable for future model versions? Require preprocessing digest parity and contract tests for every model and feature pipeline update. Maintain policy test suites and review false refusals whenever model and safety-filter versions change. Version retrieval indexes and evaluation prompts and gate releases on grounded-answer regression tests. Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. ExplanationCorrect Answer D:Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. This protects the selected approach because Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. The immediate solution is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. Other Answers A:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. B:Model instructions are probabilistic and safety guardrails need separate policy checks at input and output boundaries. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation. C:Generative AI answer quality can regress through retrieval or prompt changes even when infrastructure is healthy. The continuing safeguard required here is: Establish label-lag-aware concept-drift reviews and trigger retraining only after confirmed degradation.
Question 1050A subscription churn model's decision boundary shifts after a major change in user incentives. What should the team change to address the specific technical requirement? Evaluate model outcomes and error rates across relevant groups and apply documented responsible-AI mitigations. Configure Agent Platform Model Monitoring to alert on meaningful feature-distribution drift from a valid reference dataset. Add training-serving feature-skew checks and deploy the exact versioned preprocessing artifacts used during training. Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. ExplanationCorrect Answer D:Concept drift changes the relationship between features and targets rather than necessarily changing feature marginals. To validate the correction: Compare recent labeled metrics and calibration to a frozen prior period under comparable evaluation rules. Other Answers A:Aggregate accuracy can obscure uneven subgroup errors and biased training labels or sampling distributions. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. B:Data drift is a change in input distributions and can occur even when prediction-outcome labels are delayed. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone. C:Training-serving skew can degrade predictions without any change to the underlying fitted model weights. It does not resolve the specific failure; the targeted correction is: Monitor labeled performance and investigate concept drift using refreshed ground truth instead of input drift alone.