AI and LLMs · Middle · Question 62 of 75
How should an app handle a model answer that fails validation repeatedly?
- 1Show the invalid output to the user as is
- 2Retry forever until the output is finally valid
- 3Stop after a few retries and use a fallback or clear error
Show the answer
Correct answer: Stop after a few retries and use a fallback or clear error
Explanation
Bounded retries protect cost and latency. Log the failures; they're valuable eval cases.