179 questions
No questions match those filters.
← All questions
EvaluationMedium· A/B Testing· Experimentation· Metrics
Your team ships a new reranker and macro nDCG improves...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansRelying on a single aggregate number is insufficient to establish causality. By defining slices based on the expected influence of the reranker, you create a framework to verify that the observed improvements are actually coming from the component in question rather than external noise or unrelated system changes.