Foundation, Instruction-Tuned and Reasoning LLMs CT-GENAI Practice Questions (Page 2)

Questions 6–10

1. 6expert · hard

   A test engineer is using an LLM to generate test oracles for a complex state machine. The engineer notices that the LLM sometimes produces oracles that are logically inconsistent. The engineer wants to improve the correctness of the oracles. Which approach is most likely to help?

   • AUse a foundation LLM and ask it to generate multiple oracles and then manually select the best one.
   • BSwitch to a foundation LLM with a larger parameter count.
   • CUse an instruction-tuned LLM and provide a detailed prompt describing the state machine.
   • DUse a reasoning LLM and prompt it to think step-by-step about each state transition.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which technique is commonly used to elicit reasoning from an LLM?

   • AUsing only single-word responses
   • BChain-of-thought prompting
   • CReducing the context window
   • DIncreasing the temperature to 1.5
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A company is building a chatbot to help testers query historical test results. The chatbot needs to understand natural language questions and provide concise answers. The team has a large dataset of Q&A pairs from previous support interactions. Which approach is most appropriate?

   • AFine-tune an instruction-tuned LLM on the Q&A pairs.
   • BUse a reasoning LLM to reason about each query before answering.
   • CUse a foundation LLM and rely on its pretrained knowledge to answer queries.
   • DUse a foundation LLM with a prompt containing the entire dataset.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   What is the primary role of a foundation LLM in the context of generative AI systems?

   • ATo replace all other types of models in a generative AI system
   • BTo provide a rule-based engine for deterministic decision-making
   • CTo act as a lightweight model optimized for real-time inference on edge devices
   • DTo serve as a large-scale pretrained model that can be adapted for specialized tasks
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A QA engineer needs to generate a comprehensive set of boundary value test cases for a complex input validation function. The function has multiple interacting conditions, and the engineer wants the AI to derive the test cases logically rather than just listing examples. Which type of LLM should the engineer use?

   • AA reasoning LLM with a prompt asking it to think step-by-step through the conditions.
   • BAny LLM will produce the same result because they all use the same underlying architecture.
   • CA foundation LLM with a prompt listing the function's conditions.
   • DAn instruction-tuned LLM with a prompt asking for boundary values.
   Reveal Solution Discussion Select an answer first