Optimizing foundation modelsFull notesSummaryRingkasanStoriesPracticeFine-Tuning10 exam-style questions on this lesson.Question 1 of 10. Select two.Which TWO are benefits of fine-tuning a foundation model? (Select TWO.)AImproved accuracy on specialized tasksBRemoval of the context window limitCNo need for any training dataDReduced biases from the original training dataELower cost than prompt engineeringCheck answer0 of 2 selectedQuestion 2 of 10A company wants its virtual assistant to follow commands reliably. It has thousands of prompts paired with the ideal responses. Which fine-tuning approach fits?AInstruction tuningBTransfer learningCDomain adaptationDContinuous pretrainingCheck answerQuestion 3 of 10A team wants model outputs aligned with human values and preferences. It first does supervised training, then uses a reward model built from human rankings. Which approach is this?AInstruction tuning on prompt–response pairs onlyBRLHF (reinforcement learning from human feedback)CTransfer learning from a related taskDDomain adaptation on an industry corpusCheck answerQuestion 4 of 10A law firm trains a general model further on a large corpus of legal documents so it becomes more relevant and accurate for legal work. Which approach is this?AInstruction tuningBRLHFCDomain adaptationDModel pruningCheck answerQuestion 5 of 10A team reuses a model trained for one task as the starting point for a related task, to save training effort. Which approach is this?ADomain adaptationBContinuous pretrainingCRLHFDTransfer learningCheck answerQuestion 6 of 10A news company keeps feeding its model new articles so it stays current with new vocabulary and trends. Which approach is this?AContinuous pretrainingBTransfer learningCBatch transformDInstruction tuningCheck answerQuestion 7 of 10. Select two.Which TWO are NOT fine-tuning methods? (Select TWO.)AInstruction tuningBDomain adaptationCROUGEDModel pruningERLHFCheck answer0 of 2 selectedQuestion 8 of 10How does fine-tuning data differ from pre-training data?AFine-tuning data must be larger and broader than pre-training dataBFine-tuning data is smaller, focused, and quality-firstCThey are identical, just used at different timesDFine-tuning data must be unlabeled, like pre-training dataCheck answerQuestion 9 of 10Which step in preparing fine-tuning data does the course call paramount, because it drives specialization in the target domain?ACollecting as much random data as possibleBIncreasing the temperatureCAccurate, relevant labelingDChoosing the GPU typeCheck answerQuestion 10 of 10. Select two.Which TWO are key steps in preparing fine-tuning data? (Select TWO.)AIgnoring industry regulationsBRemoving all labelsCData curation: rigorous selection of only relevant dataDRepresentativeness and bias checkingEDeleting all feedbackCheck answer0 of 2 selectedBusiness Case7 exam-style questions on this lesson.Model Evaluation9 exam-style questions on this lesson.