Generative AI and Large Language Models (LLMs) AIE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   An online learning platform uses an LLM to generate quiz questions. They notice that some questions contain subtle gender stereotypes. What is the most effective mitigation strategy?

   • AUse prompt engineering to explicitly instruct the model to avoid stereotypes and provide examples of unbiased questions.
   • BIgnore the bias because it is inherent and cannot be changed.
   • CUse a smaller model, as smaller models are less biased.
   • DIncrease the model's temperature to introduce more randomness, which will break the stereotype patterns.
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A company wants to use AI to detect fraudulent transactions in a stream of credit card purchases. Which type of AI model is most appropriate?

   • AA discriminative model that classifies transactions as fraudulent or legitimate.
   • BA language model that reads transaction descriptions.
   • CA generative model that creates new transactions to test the system.
   • DA clustering model that groups transactions without labels.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A customer-service manager wants to use an LLM to automatically categorize incoming emails as 'billing', 'technical', or 'other'. Which capability of LLMs is most relevant?

   • AText generation, because the model can generate a new email in each category.
   • BSummarization, because the model can summarize the email to determine its category.
   • CText classification, which can be performed by prompting the model to output a category label.
   • DTranslation, because the model can translate the email into a standard format.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   What is a defining characteristic of large language models (LLMs)?

   • AThey can only process numeric data, not text
   • BThey have a small number of parameters and require minimal data
   • CThey are trained exclusively on a single author's works
   • DThey are trained on vast text corpora and have billions of parameters
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A customer-support team uses a general-purpose LLM to draft replies to routine tickets. The drafts often include plausible-sounding but incorrect refund policies. The team wants to reduce these errors without retraining the model. What should they do first?

   • ASwitch to a smaller LLM that has fewer parameters and therefore fewer hallucinations.
   • BProvide the exact refund policy text in the prompt and instruct the model to base its reply only on that text.
   • CAsk the model to generate a longer response with more detail so it can reason through the policy.
   • DFine-tune the LLM on a large dataset of historical customer tickets and refund cases.
   Reveal Solution Discussion Select an answer first