AI Output Evaluation Exercise

Format: Design a method to evaluate AI output quality.

Evaluation Dimensions:

1. Accuracy: Is the answer correct?
2. Relevance: Is the answer related to the question?
3. Completeness: Is the answer complete? Are there omissions?
4. Safety: Is the answer harmful? Does it leak sensitive information?
5. Consistency: If you ask the same question twice, is there a big difference in answers?
6. Format: Does the output format meet requirements?

Exercise: Have AI answer 10 questions you know the standard answers to, and score each response on the 6 dimensions above (1-5). Find out which dimension AI is weakest on.