ai-engineering-from-scratch-zh/phases/14-agent-engineering/12-anthropic-workflow-patterns/quiz.json at main · fancyboi999/ai-engineering-from-scratch-zh · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
{
  "lesson": "12-anthropic-workflow-patterns",
  "title": "Anthropic 的工作流模式：简单胜过复杂",
  "questions": [
    {
      "stage": "pre",
      "question": "Anthropic 如何区分工作流（workflow）和 agent？",
      "options": [
        "工作流用 embedding；agent 用工具",
        "工作流是工程师拥有的预定义图；agent 是模型拥有的动态工具调度",
        "工作流跑在 CPU 上；agent 需要 GPU",
        "工作流是无状态的；agent 是有状态的"
      ],
      "correct": 1,
      "explanation": "工作流 = 工程师拥有的预定义代码路径；agent = 模型拥有那张图。"
    },
    {
      "stage": "pre",
      "question": "支撑全部五种模式的增强型 LLM 有哪三种能力？",
      "options": [
        "搜索（检索）、工具（动作）、记忆（持久化）",
        "Embedding、微调、RAG",
        "向量、KV、图",
        "Plan、execute、reflect"
      ],
      "correct": 0,
      "explanation": "原子单元是一个接好了检索、工具和记忆的 LLM。"
    },
    {
      "stage": "check",
      "question": "下列哪一个不是 Anthropic 的五种工作流模式之一？",
      "options": [
        "Prompt chaining（prompt 链）",
        "Routing（路由）",
        "Evaluator-optimizer（评估器-优化器）",
        "Gradient distillation（梯度蒸馏）"
      ],
      "correct": 3,
      "explanation": "五种是 prompt chaining、routing、parallelization、orchestrator-workers、evaluator-optimizer。梯度蒸馏是一个训练概念。"
    },
    {
      "stage": "check",
      "question": "并行化（parallelization）有哪两种形态？",
      "options": [
        "同步和异步",
        "分区（sectioning，不同的块）和投票（voting，对同一 prompt 调用 N 次再聚合）",
        "热和冷",
        "有状态和无状态"
      ],
      "correct": 1,
      "explanation": "并行化是分区或投票；两者都扇出 N 个调用并聚合。"
    },
    {
      "stage": "check",
      "question": "Self-Refine 被泛化成了哪种工作流模式？",
      "options": [
        "Prompt chaining",
        "Routing",
        "Orchestrator-workers",
        "Evaluator-optimizer"
      ],
      "correct": 3,
      "explanation": "Evaluator-optimizer 是 Anthropic 对 Self-Refine / CRITIC 迭代模式的命名。"
    },
    {
      "stage": "post",
      "question": "按本课所说，什么时候工作流胜过 agent？",
      "options": [
        "总是",
        "在可预测、有成本边界或有合规边界的任务上，图可以被枚举和审计",
        "只在聊天时",
        "只在 GPU 上"
      ],
      "correct": 1,
      "explanation": "工作流更便宜、更易调试、可审计；当步骤可知时就选它。"
    },
    {
      "stage": "post",
      "question": "本课推荐的默认起点是什么？",
      "options": [
        "一个多 agent 框架",
        "直接 API 调用；只在持久状态、actor 并发或角色模板能值回其成本时才加框架",
        "微调模型",
        "构建一个自定义 MCTS"
      ],
      "correct": 1,
      "explanation": "Schluntz 和 Zhang：从简单开始；只在有正当理由时才增加框架复杂度。"
    }
  ]
}