iSQI CT-GenAI Practice Exam Questions

iSQI CT-GenAI Sample Question Answers

Question # 1

Who typically defines the system prompt in a testing workflow? 

A. A tester configuring the assistant 
B. End user during normal chat use 
C. CI server automatically without human input 
D. Product owner in user stories only 

Show Answer

Question # 2

An LLM prioritizes tests using likelihood X impact but ranks a trivial tooltip change above apayment failure. What defect does this MOST LIKELY show?

A. No defect; this is acceptable 
B. Reasoning error in risk calculation logic 
C. Hallucination 
D. Dataset bias toward UI features 

Show Answer

Question # 3

Which concept refers to breaking text into smaller units for processing by LLMs? 

A. Transformer 
B. Embeddings 
C. Context Window 
D. Tokenization 

Show Answer

Question # 4

Which setting can reduce variability by narrowing the sampling distribution during inference?

A. Increasing temperature 
B. Increasing learning rate 
C. Lowering temperature 
D. Using a larger context window 

Show Answer

Question # 5

Which statement BEST describes vision-language models (VLMs)? 

A. VLMs are a subset of multimodal LLMs integrating visual and textual information. 
B. VLMs are unrelated to multimodal LLMs and focus only on UI automation. 
C. VLMs are a superset of multimodal LLMs. 
D. VLMs process audio and video but not images. 

Show Answer

Question # 6

Which standard specifies requirements for managing AI systems within an organization,supporting consistent GenAI use in testing?

A. ISO/IEC 42001:2023 
B. NIST AI RMF 1.0 
C. ISO/IEC 23053:2022 
D. EU AI Act 

Show Answer

Question # 7

Consider applying the meta-prompting technique to generate automated test scripts for APItesting. You need to test a REST API endpoint that processes user registration withvalidation rules. Which one of the following prompts is BEST suited to this task?

A. Role: Act as a test automation engineer with API testing experience. | Context: You areverifying user registration that enforces field and format validation. | Instruction: Generatepytest scripts using requests for both positive (valid) and negative (invalid email, weakpassword, missing fields) cases. | Input Data: POST /api/register with validation rules foremail and password length. | Constraints: Include fixtures, clear assertions, and namingconsistent with pytest. | Output Format: Return complete Python test files.
B. Role: Act as a test automation engineer. | Context: You are creating tests for aregistration endpoint. | Instruction: Generate Python test scripts using pytest covering bothvalid and invalid inputs. | Input Data: POST /api/register with email and password. |Constraints: Follow pytest structure. | Output Format: Provide scripts.
C. Role: Act as an automation tester. | Context: You are validating an API endpoint. |Instruction: Generate Python test scripts that send POST requests and validate responses.| Input Data: User credentials. | Constraints: Include basic scenarios with asserts. | OutputFormat: Provide organized scripts.
D. Role: Act as a software engineer. | Context: You are testing registration logic. |Instruction: Create Python scripts to verify endpoint behavior. | Input Data: POST/api/register with test users. | Constraints: Add checks for status codes. | Output Format:Deliver functional scripts.

Show Answer

Question # 8

A tester uploads crafted images that steer the LLM into validating non-existent acceptance criteria. Which attack vector is this? 

A. Data poisoning 
B. Data exfiltration 
C. Request manipulation 
D. Malicious code generation 

Show Answer

Question # 9

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

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

Show Answer

Question # 10

What is a hallucination in LLM outputs? 

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

Show Answer