examination/topics/qunar-ai-fullstack/ai-structured-output/single_choice.json

{
  "topic": "ai-structured-output",
  "type": "single_choice",
  "schema_version": "1.0.0",
  "generated": "2026-09-02T00:00:00Z",
  "questions": [
    {
      "id": "sc-001",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种结构化输出方法能 100% 保证符合 Schema？",
      "explanation": "Structured Output 使用受限解码屏蔽非法 token，保证 100% 符合 Schema。",
      "source": null,
      "related": [],
      "options": {
        "A": "JSON Mode",
        "B": "Structured Output（受限解码）",
        "C": "后处理 + 重试",
        "D": "分步生成"
      },
      "answer": "B"
    },
    {
      "id": "sc-002",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种 Agent 模式最适合需要全局规划后再执行的复杂任务？",
      "explanation": "Plan-and-Execute 先制定完整计划，再逐步执行，适合需要全局规划的复杂任务。",
      "source": null,
      "related": [],
      "options": {
        "A": "ReAct",
        "B": "Reflexion",
        "C": "Plan-and-Execute",
        "D": "Multi-Agent"
      },
      "answer": "C"
    },
    {
      "id": "sc-003",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种 Agent 模式最适合步骤不确定的通用任务？",
      "explanation": "ReAct 交替进行推理和行动，灵活应对步骤不确定的通用任务。",
      "source": null,
      "related": [],
      "options": {
        "A": "Plan-and-Execute",
        "B": "Reflexion",
        "C": "Critic/Verifier",
        "D": "ReAct"
      },
      "answer": "D"
    },
    {
      "id": "sc-004",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种结构化输出方法可靠性中等、灵活性高？",
      "explanation": "JSON Mode 可靠性中等（格式对但内容可能错）、灵活性高、延迟低。",
      "source": null,
      "related": [],
      "options": {
        "A": "Structured Output",
        "B": "Function Calling",
        "C": "JSON Mode",
        "D": "分步生成"
      },
      "answer": "C"
    },
    {
      "id": "sc-005",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种 Agent 模式包含两个独立的 Agent——一个执行、一个审查？",
      "explanation": "Critic/Verifier 模式使用执行 Agent + 审查 Agent 来实现高可靠性。",
      "source": null,
      "related": [],
      "options": {
        "A": "ReAct",
        "B": "Multi-Agent",
        "C": "Critic/Verifier",
        "D": "Reflexion"
      },
      "answer": "C"
    },
    {
      "id": "sc-006",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "以下哪些结构化输出方法的可靠性为「高」？",
      "explanation": "Structured Output、Function Calling 和分步生成可靠性都为「高」。",
      "source": null,
      "related": [],
      "options": {
        "A": "JSON Mode 和后处理+重试",
        "B": "Structured Output、Function Calling 和分步生成",
        "C": "仅 Structured Output",
        "D": "所有五种方法"
      },
      "answer": "B"
    },
    {
      "id": "sc-007",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "严格的 JSON Schema 应该包含哪些约束？",
      "explanation": "最佳实践推荐使用 required、enum 约束和 additionalProperties: false。",
      "source": null,
      "related": [],
      "options": {
        "A": "additionalProperties: true，无 required 字段",
        "B": "required、enum 和 additionalProperties: false",
        "C": "仅 type 定义，无约束",
        "D": "所有字段使用通配符"
      },
      "answer": "B"
    },
    {
      "id": "sc-008",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "后处理+重试方法的可靠性等级是？",
      "explanation": "后处理+重试可靠性中等——解析失败触发重试，但不能保证正确性，延迟也不确定。",
      "source": null,
      "related": [],
      "options": {
        "A": "高",
        "B": "中",
        "C": "低",
        "D": "不确定"
      },
      "answer": "B"
    },
    {
      "id": "sc-009",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "哪种 Agent 模式涉及多个 Agent 协作，各司其职？",
      "explanation": "Multi-Agent 使用多个 Agent 协作，各自承担专门角色，适合复杂系统。",
      "source": null,
      "related": [],
      "options": {
        "A": "ReAct",
        "B": "Reflexion",
        "C": "Plan-and-Execute",
        "D": "Multi-Agent"
      },
      "answer": "D"
    },
    {
      "id": "sc-010",
      "type": "single_choice",
      "difficulty": 3,
      "tags": [
        "ai",
        "structured-output",
        "agent"
      ],
      "question": "受限解码中，每步 token 生成时会发生什么？",
      "explanation": "每步根据当前状态和 Schema 约束屏蔽非法 token，模型只能从合法 token 中采样。",
      "source": null,
      "related": [],
      "options": {
        "A": "模型生成任意 token 然后验证",
        "B": "根据当前状态和 Schema 约束屏蔽非法 token",
        "C": "模型输出完整 JSON 后校验",
        "D": "独立的验证模型对每个 token 打分"
      },
      "answer": "B"
    }
  ]
}