Google Professional Machine Learning Engineer Practice Test 56
Question 550
While reproducing a training orchestration incident in isolated staging, the following condition occurs: A CI/CD workflow always deploys the most recent checkpoint even when its validation F1 is below the previous approved baseline. The team requires an idempotent pipeline behavior and traceable evidence. Which outcome most directly demonstrates a successful, measurable fix?
- A deliberately malformed dataset is rejected before training while healthy fixtures proceed successfully.
- An intentionally weak candidate cannot deploy, whereas an approved passing model can proceed with traceable evidence.
- A deliberately incompatible payload prevents the release and a compatible fixture completes an end-to-end prediction.
- An audit can resolve model version to exact source data and processing code without guessing by timestamps.
Explanation
Correct Answer
B:An intentionally weak candidate cannot deploy, whereas an approved passing model can proceed with traceable evidence. This is specific acceptance evidence for model evaluation gate before deployment. Evaluation without a blocking release decision cannot protect production from quality regressions. It confirms the controlled correction: Require candidate evaluation against a governed baseline with an enforced acceptance threshold before promotion. This matters here because while reproducing a training orchestration incident in isolated staging, the following condition occurs: A CI/CD….
Incorrect Answers
A:A deliberately malformed dataset is rejected before training while healthy fixtures proceed successfully. This result could demonstrate success for data validation threshold gate but not for the stated model evaluation gate before deployment issue. The direct test is: An intentionally weak candidate cannot deploy, whereas an approved passing model can proceed with traceable evidence.
C:A deliberately incompatible payload prevents the release and a compatible fixture completes an end-to-end prediction. This result could demonstrate success for continuous delivery integration contract tests but not for the stated model evaluation gate before deployment issue. The direct test is: An intentionally weak candidate cannot deploy, whereas an approved passing model can proceed with traceable evidence.
D:An audit can resolve model version to exact source data and processing code without guessing by timestamps. This result could demonstrate success for model feature lineage for audit but not for the stated model evaluation gate before deployment issue. The direct test is: An intentionally weak candidate cannot deploy, whereas an approved passing model can proceed with traceable evidence.
Question 551
Within a jointly maintained Google Cloud data science workspace, the following condition occurs: A churn model joins historical feature rows to today's latest cancellation status, exposing a future outcome to prior prediction dates. Collaborators must avoid expanding data access or breaking experiment reproducibility. Which TWO selections represent the correct investigation and remediation together? Select TWO.
- Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment.
- Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags.
- Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden.
- Record input dataset versions, training runs, code revisions and output artifact links in the managed metadata and registry workflow.
- Handle the dominant key separately or repartition and salt the skewed join while validating join semantics and counts.
Explanation
Correct Answers
A:Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment. This is the targeted corrective response for point-in-time feature leakage prevention. The globally latest value may contain future information that could not have informed the original prediction. Verify the diagnosis using: Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags. Confirm success using: For every row the feature version predates its prediction cutoff and prospective holdout quality is realistic. This matters here because within a jointly maintained Google Cloud data science workspace, the following condition occurs: A churn model joins….
B:Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags. This inspects evidence relevant to point-in-time feature leakage prevention. The globally latest value may contain future information that could not have informed the original prediction. After identifying the cause, the matching correction is: Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment. This matters here because within a jointly maintained Google Cloud data science workspace, the following condition occurs: A churn model joins….
Incorrect Answers
C:Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden. This is an action or diagnostic relevant to bigquery versus dataflow preparation selection, not one of the required two steps. In this scenario verify: Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags. Then correct: Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment.
D:Record input dataset versions, training runs, code revisions and output artifact links in the managed metadata and registry workflow. This is an action or diagnostic relevant to ml metadata lineage for production model, not one of the required two steps. In this scenario verify: Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags. Then correct: Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment.
E:Handle the dominant key separately or repartition and salt the skewed join while validating join semantics and counts. This is an action or diagnostic relevant to managed spark join-key skew repair, not one of the required two steps. In this scenario verify: Inspect feature event timestamps, prediction cutoffs, feature joins and any post-cancellation administrative flags. Then correct: Rebuild offline training data with point-in-time eligible joins and remove features available only after each prediction moment.
Question 552
During a postdeployment AI safety and reliability review, the following condition occurs: A product team interprets a high feature attribution for zip code as proof that changing a customer's address would cause a credit decision to improve. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. What acceptance result verifies the intended behavior after remediation?
- For controlled entities the online and offline features match at equivalent cutoffs and quality recovers.
- Attack-bearing documents fail to redirect privileged behavior while legitimate grounded answers remain available.
- Test prompts preserve support utility without exposing real account identifiers to unauthorized model requests.
- Review documentation correctly distinguishes model influence from proven real-world causal effects.
Explanation
Correct Answer
D:Review documentation correctly distinguishes model influence from proven real-world causal effects. This is specific acceptance evidence for explainability versus causal claim. Predictive importance is not causal proof; attribution techniques describe model behavior under their assumptions. It confirms the controlled correction: Present feature attributions as model-decision associations and use proper causal analysis before making intervention claims.
Incorrect Answers
A:For controlled entities the online and offline features match at equivalent cutoffs and quality recovers. This result could demonstrate success for training-serving feature skew detection but not for the stated explainability versus causal claim issue. The direct test is: Review documentation correctly distinguishes model influence from proven real-world causal effects.
B:Attack-bearing documents fail to redirect privileged behavior while legitimate grounded answers remain available. This result could demonstrate success for prompt injection through retrieved documents but not for the stated explainability versus causal claim issue. The direct test is: Review documentation correctly distinguishes model influence from proven real-world causal effects.
C:Test prompts preserve support utility without exposing real account identifiers to unauthorized model requests. This result could demonstrate success for sensitive identifiers leaked to llm prompts but not for the stated explainability versus causal claim issue. The direct test is: Review documentation correctly distinguishes model influence from proven real-world causal effects.
Question 553
During a controlled replay of the data and model handoff to a second engineering team, the following condition occurs: An analyst runs a long-lived custom Apache Beam streaming service to compute a simple daily SQL join and aggregate over tables already in BigQuery. The handoff must preserve privacy, lineage and testable behavior. What should the engineer inspect before selecting a correction?
- Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden.
- Inspect feature quantiles, physical validity bounds, missingness and training-only fit statistics against independent validation data.
- Inspect the labeling payload schema, data classification, annotator roles and fields needed to complete the labeling task.
- Inspect random seeds, sample ordering, framework determinism settings and environment versions recorded per run.
Explanation
Correct Answer
A:Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden. This inspects evidence relevant to bigquery versus dataflow preparation selection. Choosing a managed SQL engine for warehouse-resident batch joins avoids unnecessary always-on stream processing. After identifying the cause, the matching correction is: Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA.
Incorrect Answers
B:Inspect feature quantiles, physical validity bounds, missingness and training-only fit statistics against independent validation data. This examines outlier handling before feature standardization rather than the specific mechanism reported here. It is not the decisive first check for bigquery versus dataflow preparation selection. The relevant evidence is: Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden.
C:Inspect the labeling payload schema, data classification, annotator roles and fields needed to complete the labeling task. This examines data minimization for labeling inputs rather than the specific mechanism reported here. It is not the decisive first check for bigquery versus dataflow preparation selection. The relevant evidence is: Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden.
D:Inspect random seeds, sample ordering, framework determinism settings and environment versions recorded per run. This examines experiment random seed and data order variance rather than the specific mechanism reported here. It is not the decisive first check for bigquery versus dataflow preparation selection. The relevant evidence is: Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden.
Question 554
During a staged restoration test at a secondary serving deployment, the following condition occurs: A new image classifier has only offline test results, yet engineers plan to replace every customer prediction immediately after deployment. The team must prove an actual user-visible improvement without weakening access controls. Which TWO selections form an appropriate before-and-after verification pair? Select TWO.
- Identical sample text yields matching token IDs and model outputs in offline and online environments.
- Inspect traffic-splitting capabilities, release guardrails, rollback triggers and representative online metrics.
- Review payload size distributions, request duration, response timeout policies and tolerance for deferred completion.
- Only a limited portion initially uses the candidate and a simulated regression triggers a tested rollback.
- Previously missed customer IDs retrieve the intended feature values without collapsing distinct leading-zero identifiers.
Explanation
Correct Answers
B:Inspect traffic-splitting capabilities, release guardrails, rollback triggers and representative online metrics. This inspects evidence relevant to canary gradual traffic rollout. Progressive traffic exposure reduces blast radius compared with instant replacement of all production serving. After identifying the cause, the matching correction is: Roll out a small canary percentage first, monitor error and quality guardrails and expand only after acceptance.
D:Only a limited portion initially uses the candidate and a simulated regression triggers a tested rollback. This is specific acceptance evidence for canary gradual traffic rollout. Progressive traffic exposure reduces blast radius compared with instant replacement of all production serving. It confirms the controlled correction: Roll out a small canary percentage first, monitor error and quality guardrails and expand only after acceptance.
Incorrect Answers
A:Identical sample text yields matching token IDs and model outputs in offline and online environments. This relates to inference preprocessing tokenization mismatch but does not establish the requested before-and-after evidence. The actual diagnostic is: Inspect traffic-splitting capabilities, release guardrails, rollback triggers and representative online metrics. The specific acceptance result is: Only a limited portion initially uses the candidate and a simulated regression triggers a tested rollback.
C:Review payload size distributions, request duration, response timeout policies and tolerance for deferred completion. This relates to online prediction timeout and batch fallback but does not establish the requested before-and-after evidence. The actual diagnostic is: Inspect traffic-splitting capabilities, release guardrails, rollback triggers and representative online metrics. The specific acceptance result is: Only a limited portion initially uses the candidate and a simulated regression triggers a tested rollback.
E:Previously missed customer IDs retrieve the intended feature values without collapsing distinct leading-zero identifiers. This relates to feature online entity key mismatch but does not establish the requested before-and-after evidence. The actual diagnostic is: Inspect traffic-splitting capabilities, release guardrails, rollback triggers and representative online metrics. The specific acceptance result is: Only a limited portion initially uses the candidate and a simulated regression triggers a tested rollback.
Question 555
During a staged restoration test at a secondary serving deployment, the following condition occurs: An online transformer processes one request per GPU kernel launch and throughput stays low despite steady high-frequency traffic. The team must prove an actual user-visible improvement without weakening access controls. Which result is the strongest postchange evidence of success?
- The selected service meets p95 latency and payload requirements with lower idle cost and reliable scaling.
- A test client on the approved network succeeds while an external probe remains blocked.
- Peak requests trigger additional serving capacity and tail latency recovers without repeated scale oscillations.
- Requests per second increase while tail latency stays inside the service objective.
Explanation
Correct Answer
D:Requests per second increase while tail latency stays inside the service objective. This is specific acceptance evidence for dynamic batch serving for gpu saturation. Bundling compatible requests can amortize GPU launch overhead, but excessive batch waiting raises response times. It confirms the controlled correction: Enable compatible dynamic batching with a bounded wait time and benchmark throughput against p95 latency.
Incorrect Answers
A:The selected service meets p95 latency and payload requirements with lower idle cost and reliable scaling. This result could demonstrate success for cloud run versus managed inference endpoint but not for the stated dynamic batch serving for gpu saturation issue. The direct test is: Requests per second increase while tail latency stays inside the service objective.
B:A test client on the approved network succeeds while an external probe remains blocked. This result could demonstrate success for private vpc routing failure for serving but not for the stated dynamic batch serving for gpu saturation issue. The direct test is: Requests per second increase while tail latency stays inside the service objective.
C:Peak requests trigger additional serving capacity and tail latency recovers without repeated scale oscillations. This result could demonstrate success for throughput-based autoscaling versus cpu metric but not for the stated dynamic batch serving for gpu saturation issue. The direct test is: Requests per second increase while tail latency stays inside the service objective.
Question 556
During model-training validation on a Google Cloud experimental project, the following condition occurs: A customer-service model correctly follows general instructions but lacks newly updated internal warranty details, so developers schedule daily weight fine-tuning. The change should preserve a meaningful independent model-quality comparison. Which targeted engineering change most directly addresses the failure?
- Tune supported distributed batch sizing, communication strategy or placement after measuring the bottleneck.
- Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
- Evaluate ARIMA_PLUS or an appropriate time-series forecasting model before adopting an expensive generative approach.
- Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
Explanation
Correct Answer
B:Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation. This is the targeted corrective response for foundation model prompt versus fine-tuning. External reference updates need fresh context; changing foundation weights for every factual revision is inefficient. Verify the diagnosis using: Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage. Confirm success using: New warranty rules appear correctly after the index update without unnecessary retraining and sources remain traceable. This matters here because during model-training validation on a Google Cloud experimental project, the following condition occurs: A….
Incorrect Answers
A:Tune supported distributed batch sizing, communication strategy or placement after measuring the bottleneck. This action addresses distributed gradient communication bottleneck, not the failure in the described foundation model prompt versus fine-tuning case. The targeted correction is: Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
C:Evaluate ARIMA_PLUS or an appropriate time-series forecasting model before adopting an expensive generative approach. This action addresses forecast trend and seasonality versus llm, not the failure in the described foundation model prompt versus fine-tuning case. The targeted correction is: Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
D:Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses. This action addresses constrained generative json outputs, not the failure in the described foundation model prompt versus fine-tuning case. The targeted correction is: Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
Question 557
During a production ML inference release on Google Cloud, the following condition occurs: A multi-class model returns numeric class indices, but an API gateway uses a stale alphabetical mapping and mislabels predictions. The service must meet its agreed quality and latency requirements. Which underlying engineering principle explains why the original approach fails?
- A cloud-only inference architecture cannot satisfy hard offline and low-latency edge requirements.
- Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference.
- CPU usage can be weakly related to accelerator or queue-bound inference workloads.
- Embedding matrices interpret token IDs according to one vocabulary; mismatched preprocessing changes input meaning.
Explanation
Correct Answer
B:Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference. This explains the mechanism behind custom postprocessing label mapping drift and why the described failure is possible. A discriminating investigation is: Inspect trained class-to-index metadata, endpoint output tensors and gateway mapping version. The corresponding repair is: Ship the class vocabulary with the model artifact and version or validate the output adapter against it.
Incorrect Answers
A:A cloud-only inference architecture cannot satisfy hard offline and low-latency edge requirements. This principle describes edge versus cloud inference network constraints, which is not the cause of the observed custom postprocessing label mapping drift failure. The controlling explanation is: Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference.
C:CPU usage can be weakly related to accelerator or queue-bound inference workloads. This principle describes throughput-based autoscaling versus cpu metric, which is not the cause of the observed custom postprocessing label mapping drift failure. The controlling explanation is: Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference.
D:Embedding matrices interpret token IDs according to one vocabulary; mismatched preprocessing changes input meaning. This principle describes inference preprocessing tokenization mismatch, which is not the cause of the observed custom postprocessing label mapping drift failure. The controlling explanation is: Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference.
Question 558
Within a jointly maintained Google Cloud data science workspace, the following condition occurs: A team uses manual spreadsheet rows for hyperparameter results and loses connections between training metrics and produced checkpoint artifacts. Collaborators must avoid expanding data access or breaking experiment reproducibility. Which underlying engineering principle explains why the original approach fails?
- Metrics from different data distributions are not directly comparable evidence of model improvement.
- LLM-based scoring can be sensitive to phrasing, order and model preferences; evaluation needs explicit calibration.
- Trial metrics and artifact provenance must be linked to support valid experiment selection and release audits.
- Long-lived secrets copied into notebooks or Git can be reused outside intended workloads and must be rotated.
Explanation
Correct Answer
C:Trial metrics and artifact provenance must be linked to support valid experiment selection and release audits. This explains the mechanism behind pipeline experiment versus notebook-only tracking and why the described failure is possible. A discriminating investigation is: Inspect pipeline execution IDs, experiment tracking records, artifact URIs and the link between metrics and saved models. The corresponding repair is: Use Agent Platform Experiments and pipeline metadata to track each trial's parameters, measurements and output artifacts.
Incorrect Answers
A:Metrics from different data distributions are not directly comparable evidence of model improvement. This principle describes experiment metric comparison with mismatched splits, which is not the cause of the observed pipeline experiment versus notebook-only tracking failure. The controlling explanation is: Trial metrics and artifact provenance must be linked to support valid experiment selection and release audits.
B:LLM-based scoring can be sensitive to phrasing, order and model preferences; evaluation needs explicit calibration. This principle describes llm-as-a-judge rubric reliability, which is not the cause of the observed pipeline experiment versus notebook-only tracking failure. The controlling explanation is: Trial metrics and artifact provenance must be linked to support valid experiment selection and release audits.
D:Long-lived secrets copied into notebooks or Git can be reused outside intended workloads and must be rotated. This principle describes workbench credentials embedded in repository, which is not the cause of the observed pipeline experiment versus notebook-only tracking failure. The controlling explanation is: Trial metrics and artifact provenance must be linked to support valid experiment selection and release audits.
Question 559
During model-training validation on a Google Cloud experimental project, the following condition occurs: An LLM occasionally emits natural-language introductions before an API-required JSON object, causing strict parsers to reject otherwise correct answers. The change should preserve a meaningful independent model-quality comparison. Select TWO: the decisive diagnostic step and the targeted repair for this specific issue.
- Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
- Examine model governance constraints, feature sensitivity, explanation fidelity and performance of simpler interpretable baselines.
- Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts.
- Use appropriately sharded data, parallel decode, caching or prefetching within memory limits to feed the accelerator.
- Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
Explanation
Correct Answers
C:Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. This inspects evidence relevant to constrained generative json outputs. Prompt-only requests for JSON may not guarantee type validity; interface contracts need machine-verifiable constraints. After identifying the cause, the matching correction is: Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
E:Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses. This is the targeted corrective response for constrained generative json outputs. Prompt-only requests for JSON may not guarantee type validity; interface contracts need machine-verifiable constraints. Verify the diagnosis using: Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. Confirm success using: A batch of representative requests yields valid schema-conforming objects and rejects invalid types explicitly.
Incorrect Answers
A:Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation. This is an action or diagnostic relevant to foundation model prompt versus fine-tuning, not one of the required two steps. In this scenario verify: Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. Then correct: Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
B:Examine model governance constraints, feature sensitivity, explanation fidelity and performance of simpler interpretable baselines. This is an action or diagnostic relevant to interpretability-constrained model baseline, not one of the required two steps. In this scenario verify: Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. Then correct: Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
D:Use appropriately sharded data, parallel decode, caching or prefetching within memory limits to feed the accelerator. This is an action or diagnostic relevant to accelerator starvation from slow input pipeline, not one of the required two steps. In this scenario verify: Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. Then correct: Use supported structured output and schema validation with graceful retry or rejection for nonconforming responses.
Question 560
While preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs: A weeklong custom training job uses preemptible capacity but saves checkpoints only to a temporary local container filesystem. Engineers need an auditable, resource-aware correction to the issue. Which result is the strongest postchange evidence of success?
- A controlled interruption resumes near the saved step without restarting training from zero.
- Domain accuracy improves while broad language capability remains within the agreed retention threshold.
- Auditors can reproduce feature influence evidence while holdout quality remains within the approved threshold.
- The training manifest contains traceable permitted sources and safety tests detect known prohibited outputs.
Explanation
Correct Answer
A:A controlled interruption resumes near the saved step without restarting training from zero. This is specific acceptance evidence for training checkpoint persistence across interruptions. Preemptible execution can end abruptly; local-only state is not a durable recovery mechanism. It confirms the controlled correction: Persist versioned checkpoints in durable Cloud Storage and make the training code resume optimizer and model state.
Incorrect Answers
B:Domain accuracy improves while broad language capability remains within the agreed retention threshold. This result could demonstrate success for fine-tuning catastrophic forgetting but not for the stated training checkpoint persistence across interruptions issue. The direct test is: A controlled interruption resumes near the saved step without restarting training from zero.
C:Auditors can reproduce feature influence evidence while holdout quality remains within the approved threshold. This result could demonstrate success for interpretability-constrained model baseline but not for the stated training checkpoint persistence across interruptions issue. The direct test is: A controlled interruption resumes near the saved step without restarting training from zero.
D:The training manifest contains traceable permitted sources and safety tests detect known prohibited outputs. This result could demonstrate success for fine-tuning image generator data rights but not for the stated training checkpoint persistence across interruptions issue. The direct test is: A controlled interruption resumes near the saved step without restarting training from zero.