Identify Hallucinations, Reasoning Errors and Biases in LLM Output CT-GENAI Practice Questions (Page 1)

Questions 1–5

1. 1foundation · easy

   How can the architecture of an LLM contribute to hallucinations?

   • AThe model's architecture is deterministic, so it always produces the same output for a given input.
   • BThe model is designed to always provide an answer, even when it lacks sufficient information.
   • CThe model's architecture forces it to repeat training data verbatim.
   • DThe model's architecture prevents it from accessing any external knowledge during inference.
   Reveal Solution Discussion Select an answer first
2. 2application · medium

   An LLM is used to generate test cases for a job application portal. The team notices that the LLM's test cases often assume the applicant is male and has a technical background, even when the prompt is neutral. The team wants to identify the type of bias. Which type of bias is this?

   • AAlgorithmic amplification
   • BCultural bias
   • CConfirmation bias
   • DGender bias
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   In the context of LLM outputs, what is a bias?

   • AA type of hallucination
   • BA systematic preference for certain types of information or groups over others
   • CA deliberate attempt by the model to mislead users
   • DA random error in the model's output
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which of the following is an example of a reasoning error, not a hallucination?

   • AThe model says that the sky is blue.
   • BThe model says 'I am not sure about that.'
   • CThe model states that the capital of Australia is Sydney.
   • DThe model correctly states that all dogs are animals and concludes that all animals are dogs.
   Reveal Solution Discussion Select an answer first
6. 5application · medium

   A test team uses an LLM to generate test scripts for a critical system. They have encountered hallucinations, reasoning errors, and biases in the outputs. The team wants to implement a comprehensive mitigation strategy. Which approach is most effective?

   • ARely on the LLM's self-assessment to ensure quality
   • BUse a single, more powerful LLM to generate all test scripts
   • CLimit the LLM's output to simple test cases to avoid errors
   • DImplement a multi-layered validation process including fact-checking, logic review, and fairness checks
   Reveal Solution Discussion Select an answer first