Question 28
Domain 3: Applications of Foundation ModelsIn which stage of the generative AI model lifecycle are tests performed to examine the model's accuracy?
Correct answer: D
Explanation
Evaluation is the lifecycle stage where model performance is measured. The source table lists “Evaluation” as the stage for “Performance and quality scoring,” and the text says SageMaker Pipelines can “gate deployments on evaluation thresholds,” showing tests are run there to assess accuracy.
Why each option is right or wrong
A. Deployment
Deployment serves inference APIs and endpoints, not accuracy testing.
B. Data selection
Data selection chooses and prepares datasets before training and evaluation.
C. Fine-tuning
Fine-tuning adapts a model on private data; evaluation measures its quality afterward.
D. Evaluation
The evaluation stage is where model performance is measured against test or validation data, and the source table explicitly places “performance and quality scoring” under Evaluation. Under the SageMaker Pipelines description, deployment can be gated on evaluation thresholds, which means the accuracy test is run before release, not during training or deployment.