Question 2: CCAR-P: Claude Certified Architect - Professional
You are integrating human review into a high-volume classification pipeline where reviewing every output is infeasible. Which sampling strategy best balances throughput with quality oversight?
Nội dung câu hỏi
You are integrating human review into a high-volume classification pipeline where reviewing every output is infeasible. Which sampling strategy best balances throughput with quality oversight?
Các lựa chọn
Đáp án được giữ gọn theo nhãn A, B, C, D trong phần bình chọn tương tác.
- A. no sampling at all, relying entirely on user-submitted complaints to surface quality and safety issues after they have already affected end users at scale
- B. risk-stratified sampling that reviews all low-confidence and high-impact outputs and a smaller random sample of high-confidence routine outputs — đáp án hiện tại
- C. inverse sampling that reviews only high-confidence routine outputs and skips all low-confidence and high-impact outputs
- D. universal sampling that reviews every output regardless of confidence level or throughput impact, treating all outputs as equally requiring human review
Answer researched independently against supporting documentation; not a published source answer key. Independent reasoning (not a published answer key): Prioritize risky or uncertain outputs while randomly auditing routine outputs to detect overlooked failures. References: https://www.anthropic.com/research/demystifying-evals-for-ai-agents https://www.anthropic.com/research/trustworthy-agents
Cộng đồng
0 bình luận công khai. Tên thành viên được ẩn một phần.