AI and LLMs · Senior · Question 68 of 75
How do you avoid an eval suite that only tests what the team already thought of?
- 1Let the model invent all new cases each week
- 2Feed it production failures, red-team findings and feedback
- 3Keep the same cases forever so scores stay comparable
Show the answer
Correct answer: Feed it production failures, red-team findings and feedback
Explanation
Evals should evolve with real usage while a stable core keeps results comparable over time.