ai-engineering-from-scratch-zh/phases/19-capstone-projects/09-code-migration-agent/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": "09-code-migration-agent",
  "title": "毕业项目 09 —— 代码迁移 agent（仓库级语言 / 运行时升级）",
  "questions": [
    {
      "stage": "pre",
      "question": "为什么流水线要把确定性底座与 agent 层结合，而不是只用其中一种？",
      "options": [
        "OpenRewrite 或 libcst 能安全且廉价地处理 70-80% 的机械式重写，把模糊的长尾留给 agent",
        "单靠 agent 比配方（recipe）更快",
        "确定性配方比 LLM 更慢",
        "只有构建系统才需要确定性"
      ],
      "correct": 0,
      "explanation": ""
    },
    {
      "stage": "pre",
      "question": "流水线用什么信号作为一次迁移成功的真值？",
      "options": [
        "agent 按细则自评",
        "评审者在 PR 上批准",
        "diff 大小低于硬性上限",
        "沙箱中 CI 通过（变绿），且覆盖率回退不超过一个很小的阈值"
      ],
      "correct": 3,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "agent 循环为每个仓库强制哪些预算上限？",
      "options": [
        "时间和成本无限制；仅在出错时中止",
        "无轮次上限，100 美元封顶",
        "1 小时挂钟时间和 100 轮",
        "30 分钟挂钟时间、8 美元成本和 20 个 agent 轮次"
      ],
      "correct": 3,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "当迁移后覆盖率下降超过约 2% 时，会触发什么门禁？",
      "options": [
        "绕过评审者",
        "该仓库被归入 coverage_regression 失败类别，而不是开一个干净的 PR",
        "确定性底座重放它的配方",
        "agent 自动强制推送一个修复"
      ],
      "correct": 1,
      "explanation": ""
    },
    {
      "stage": "check",
      "question": "为什么失败分类法（failure taxonomy）被当作交付物，而非附带产物？",
      "options": [
        "它取代了测试套件",
        "它按类别归类失败的仓库，以便未来的配方作者能针对头部失败模式",
        "它是 GitHub 分支保护所要求的",
        "它满足一项合规清单"
      ],
      "correct": 1,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "本毕业项目针对 Java 8 到 17 的迁移瞄准哪个公开基准？",
      "options": [
        "亚马逊的 MigrationBench",
        "SWE-bench Pro",
        "MMLU-Pro",
        "ViDoRe v3"
      ],
      "correct": 0,
      "explanation": ""
    },
    {
      "stage": "post",
      "question": "agent 集成细则衡量修复分布的什么？",
      "options": [
        "源文件的分词速度",
        "每个仓库的强制推送次数",
        "被固定（pin）的依赖数量",
        "由 OpenRewrite 处理与由 agent 层编写的修复各占的比例"
      ],
      "correct": 3,
      "explanation": ""
    }
  ]
}