Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ
Xây dựng bài kiểm tra thư viện nhắc nhở nhóm
Kiểm tra thiết kế mục nhập thư viện nhắc nhở, các biến được đặt tên, quyền sở hữu và phiên bản, đánh giá đại diện và giới hạn đối với các yêu cầu sử dụng lại.
Câu hỏi 1 của 8
What problem does a team prompt library address?
Xem trước mọi câu hỏi8 câu hỏi
Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.
- What problem does a team prompt library address?
- It guarantees correct answers from any model
- It makes useful task instructions and lessons reusable across people
- It removes the need to check outputs
- It automatically updates every model
- Which details help a teammate use a prompt template consistently?
- Private credentials for connected services
- Only the author’s name
- Task, intended users, required inputs, output shape, tested tool, and known limits
- The longest possible instruction text
- Why should variable fields have clear names and requirements?
- To make output identical across different models
- To avoid any need for task instructions
- So users can identify what changes between runs and validate the supplied values
- To make every prompt contain more words
- How should a team check a prompt edit?
- Change the prompt and judge it using only one unrelated output
- Assume a longer prompt is better
- Remove its version history
- Compare old and new versions on the same representative test cases
- When should a team review or retire a library entry?
- Only when its original author leaves
- When the model, tool, policy, source material, or task changes
- Never, because saved instructions do not age
- After every single use regardless of issue
- What does a saved prompt establish about its generated output?
- That no source material is needed
- That the output is compliant
- That the output is accurate
- Only that the prompt is available for reuse; the result still needs appropriate checking
- Which current API guidance does OpenAI give for production prompts?
- Never version prompt content
- Assume prompt behavior cannot change
- Treat prompt text as application code and review/test changes with the product
- Store all production prompts only in a public marketplace
- Which sign best supports saying a prompt revision improved results?
- A maintainer liked its phrasing
- It performs better against the same defined task criteria and test cases
- It uses a newer model name
- The revised prompt is longer