1. What is Data drift ? Junior A. a change in the relationship between inputs and the target, degrading model accuracy over time B. a central system that serves consistent, versioned features to both training and online inference C. a change in the input feature distribution from what the model was trained on D. the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile
2. Which term means: "a change in the input feature distribution from what the model was trained on"? Junior A. Population Stability Index (PSI) B. Training-serving skew C. Concept drift D. Data drift
3. Which statement is correct? Junior A. Data drift — a change in the input feature distribution from what the model was trained on B. Data drift — tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback C. Data drift — a mismatch between features computed in training and those at serving time, hurting live accuracy D. Data drift — a metric quantifying how much a feature or score distribution has shifted between two periods
4. What is Concept drift ? Junior A. a metric quantifying how much a feature or score distribution has shifted between two periods B. the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile C. a change in the input feature distribution from what the model was trained on D. a change in the relationship between inputs and the target, degrading model accuracy over time
5. Which term means: "a change in the relationship between inputs and the target, degrading model accuracy over time"? Junior A. Concept drift B. Population Stability Index (PSI) C. Ground-truth lag D. Model performance monitoring
6. Which statement is correct? Junior A. Concept drift — tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback B. Concept drift — a central system that serves consistent, versioned features to both training and online inference C. Concept drift — a metric quantifying how much a feature or score distribution has shifted between two periods D. Concept drift — a change in the relationship between inputs and the target, degrading model accuracy over time
7. What is Training-serving skew ? Mid A. a metric quantifying how much a feature or score distribution has shifted between two periods B. a change in the input feature distribution from what the model was trained on C. a mismatch between features computed in training and those at serving time, hurting live accuracy D. a central system that serves consistent, versioned features to both training and online inference
8. Which term means: "a mismatch between features computed in training and those at serving time, hurting live accuracy"? Mid A. Population Stability Index (PSI) B. Ground-truth lag C. Model performance monitoring D. Training-serving skew
9. Which statement is correct? Mid A. Training-serving skew — a mismatch between features computed in training and those at serving time, hurting live accuracy B. Training-serving skew — a metric quantifying how much a feature or score distribution has shifted between two periods C. Training-serving skew — a change in the input feature distribution from what the model was trained on D. Training-serving skew — a change in the relationship between inputs and the target, degrading model accuracy over time
10. What is Feature store ? Mid A. tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback B. a change in the input feature distribution from what the model was trained on C. a change in the relationship between inputs and the target, degrading model accuracy over time D. a central system that serves consistent, versioned features to both training and online inference
11. Which term means: "a central system that serves consistent, versioned features to both training and online inference"? Mid A. Concept drift B. Feature store C. Data drift D. Training-serving skew
12. Which statement is correct? Mid A. Feature store — a change in the input feature distribution from what the model was trained on B. Feature store — a mismatch between features computed in training and those at serving time, hurting live accuracy C. Feature store — a change in the relationship between inputs and the target, degrading model accuracy over time D. Feature store — a central system that serves consistent, versioned features to both training and online inference
13. What is Population Stability Index (PSI) ? Mid A. tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback B. a metric quantifying how much a feature or score distribution has shifted between two periods C. the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile D. a central system that serves consistent, versioned features to both training and online inference
14. Which term means: "a metric quantifying how much a feature or score distribution has shifted between two periods"? Mid A. Feature store B. Population Stability Index (PSI) C. Model performance monitoring D. Data drift
15. Which statement is correct? Mid A. Population Stability Index (PSI) — tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback B. Population Stability Index (PSI) — a metric quantifying how much a feature or score distribution has shifted between two periods C. Population Stability Index (PSI) — a change in the input feature distribution from what the model was trained on D. Population Stability Index (PSI) — a mismatch between features computed in training and those at serving time, hurting live accuracy
16. What is Ground-truth lag ? Senior A. the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile B. a mismatch between features computed in training and those at serving time, hurting live accuracy C. a metric quantifying how much a feature or score distribution has shifted between two periods D. a central system that serves consistent, versioned features to both training and online inference
17. Which term means: "the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile"? Senior A. Concept drift B. Ground-truth lag C. Population Stability Index (PSI) D. Training-serving skew
18. Which statement is correct? Senior A. Ground-truth lag — the delay before true labels arrive, which forces proxy metrics to detect degradation meanwhile B. Ground-truth lag — tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback C. Ground-truth lag — a metric quantifying how much a feature or score distribution has shifted between two periods D. Ground-truth lag — a central system that serves consistent, versioned features to both training and online inference
19. What is Model performance monitoring ? Senior A. a mismatch between features computed in training and those at serving time, hurting live accuracy B. tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback C. a change in the input feature distribution from what the model was trained on D. a central system that serves consistent, versioned features to both training and online inference
20. Which term means: "tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback"? Senior A. Model performance monitoring B. Feature store C. Population Stability Index (PSI) D. Ground-truth lag
21. Which statement is correct? Senior A. Model performance monitoring — a central system that serves consistent, versioned features to both training and online inference B. Model performance monitoring — tracking live accuracy, precision/recall or business KPIs to trigger retraining or rollback C. Model performance monitoring — a mismatch between features computed in training and those at serving time, hurting live accuracy D. Model performance monitoring — a change in the input feature distribution from what the model was trained on