Optimizing foundation modelsFull notesSummaryRingkasanStoriesPracticeEvaluate Results6 exam-style questions on this lesson.Question 1 of 6A team wants to compare two versions of its model objectively on accuracy, speed, and scalability. Which evaluation approach fits?ABenchmark datasetsBCustomer interviews onlyCGuardrailsDHuman evaluationCheck answerQuestion 2 of 6A team wants to know whether responses feel appropriate for the context and whether users enjoy the experience. Which approach fits?ABLEUBHuman evaluationCPerplexityDBenchmark datasetsCheck answerQuestion 3 of 6Who should write the questions and reference answers for a benchmark dataset about a company's documents?ARandomly selected customersBThe billing teamCSubject matter experts (SMEs)DThe model itselfCheck answerQuestion 4 of 6A team wants to score thousands of model answers against the benchmark answers automatically. Which approach fits?AAsk every employee to grade a share of themBUse a confusion matrix of right and wrong answersCUse SMOTE to balance the right and wrong answersDLLM as a judge, comparing answers to the benchmarkCheck answerQuestion 5 of 6What is the recommended combined evaluation approach from the telecom case?ABenchmarks before production, human ratings afterBSkip evaluation and rely on production monitoringCUse only human evaluation, before productionDUse only benchmarks, before and after launchCheck answerQuestion 6 of 6Which is a characteristic of benchmark datasets?AOnly measure creativityBQuantitative and objectiveCQualitative and subjectiveDOnly usable after deploymentCheck answerAgents7 exam-style questions on this lesson.Business Case7 exam-style questions on this lesson.