ISQI ISTQB Certified Tester Testing with Generative AI (CT-GenAI) v1.0 認定 CT-GenAI 試験問題:

1. What is a hallucination in LLM outputs?

A) A logical mistake in multi-step deduction
B) A transient network failure during inference
C) A systematic preference learned from data
D) Generation of factually incorrect content for the task

2. Which technique MOST directly reduces hallucinations by grounding the model in project realities?

A) Use longer temperature settings
B) Rely on generic examples only
C) Randomize prompts each run
D) Provide detailed context

3. The model flags anomalies in logs and also proposes partitions for input validation tests. Which metrics BEST evaluate these two outcomes together?

A) Accuracy for anomaly detection and Precision for coverage of valid/invalid partitions
B) Time efficiency for anomaly detection and accuracy for coverage of valid/invalid partitions
C) Precision for anomaly identification and recall for coverage of valid/invalid partitions
D) Diversity for anomaly identification and precision for partitions

4. Which factor MOST influences the overall energy consumption of a Generative AI model used in software testing tasks?

A) The type of cloud platform affects processing speed but not total energy draw
B) The location of the data center determines model bias and accuracy levels
C) The duration of user sessions primarily affects latency but not power efficiency
D) The number of tokens processed directly determines the carbon intensity of each query

5. What does an embedding represent in an LLM?

A) A set of test cases for validation
B) Logical rules for reasoning
C) Numerical vectors capturing semantic relationships
D) Tokens grouped into context windows

質問と回答：