AI and LLMs · Junior · Question 33 of 75
Why is testing an LLM feature harder than testing normal code?
- 1Outputs vary, so exact string comparisons often don't work
- 2LLM output is always identical, so tests are pointless
- 3LLM features can't be called from automated tests at all
Show the answer
Correct answer: Outputs vary, so exact string comparisons often don't work
Explanation
Tests check properties instead: valid JSON, required fields, length limits, contains or excludes certain facts, or a judged quality score.