Metrics for GenAI results · Question 28 of 60
A tester runs a prompt once, finds the result very good and declares the prompt finished. What is problematic about this?
1 answer is correct. Medium Created with AI
Show answer
Answer
- B A single run is not a reliable basis for non-deterministic models
Explanation
The same model may answer less well next time. Only several runs show how good the prompt is on average and how stable it is.