Guide-linked quiz · Medium Level
AI Test Generation Quiz
Assess test-oracle limits, characterization tests, symbolic and LLM test generation, fuzzing, and coverage guidance. Check practical limits.
Question 1 of 8
What is the 'oracle problem' in AI test generation?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- What is the 'oracle problem' in AI test generation?
- Knowing the correct expected output for a given input
- Generating enough random inputs to run the program
- Connecting the test runner to a database
- Choosing which programming language to test in
- What does a 'characterization test' lock in?
- The fastest possible runtime
- The current behavior of the code as a regression baseline
- The original author's coding style
- A formal mathematical proof of correctness
- What programming language did the TestPilot study implement and evaluate its LLM test-generation approach for?
- Rust
- Fortran
- JavaScript
- SQL only
- A key risk of accepting AI-generated tests without review is that they may:
- Run too slowly to be useful
- Delete the source code automatically
- Only work on weekends
- Assert and freeze existing bugs as if they were correct
- How do LLM-based test assistants like Copilot typically produce a test?
- By predicting test code from the function and surrounding context
- By formally proving the code is correct
- By measuring CPU temperature
- By recording the developer's screen
- Coverage-guided fuzzing steers random inputs using:
- User star ratings
- Code instrumentation that measures which branches were hit
- The phase of the moon
- Only the file name
- What did the TestPilot evaluation report as its median statement and branch coverage?
- 100% statement and branch coverage on every package
- 0% statement coverage and 52.8% branch coverage
- 70.2% statement coverage and 52.8% branch coverage across its studied packages
- The paper did not measure code coverage
- What does mutation testing assess about a test suite?
- Whether every test uses the same name
- Whether a characterization test preserves all current behavior
- Whether the fuzzer can reach a branch without running tests
- Whether tests detect small, deliberate changes made to the program
Keep testing yourself
More quizzes picked for your level.