Request

{
  "type": "compare",
  "evaluationModelKey": "deepseek",
  "mode": {
    "functionMode": "basic",
    "subMode": "system"
  },
  "focus": {
    "content": "即便 reference 看上去更完整、更自然，只要 prompt 改动造成字段名或外层结构变化，就应把 target 相对 baseline 判为回退。",
    "source": "system",
    "priority": "highest"
  },
  "target": {
    "workspacePrompt": "你是电商商品信息抽取助手。\n输出一个 JSON 对象，字段改为 product_name, buyer_highlights, cautions。\n为了后续扩展，把三个字段统一包在 payload 里。\n不要解释。",
    "referencePrompt": "你是电商商品信息抽取助手。\n只输出 JSON 对象，字段必须为 title, selling_points, cautions。\n不要改字段名，不要添加外层包裹对象，不要解释。"
  },
  "testCases": [
    {
      "id": "tc-1",
      "input": {
        "kind": "text",
        "label": "商品文案",
        "content": "便携手冲咖啡壶，容量 600ml，适合露营和办公室使用，主打双层不锈钢保温，注意不支持电磁炉直火加热。"
      }
    }
  ],
  "snapshots": [
    {
      "id": "a",
      "label": "A",
      "testCaseId": "tc-1",
      "promptRef": {
        "kind": "workspace",
        "label": "Workspace"
      },
      "promptText": "你是电商商品信息抽取助手。\n输出一个 JSON 对象，字段改为 product_name, buyer_highlights, cautions。\n为了后续扩展，把三个字段统一包在 payload 里。\n不要解释。",
      "output": "{\"payload\":{\"product_name\":\"便携手冲咖啡壶\",\"buyer_highlights\":[\"600ml 容量\",\"适合露营和办公室\",\"双层不锈钢保温\"],\"cautions\":[\"不支持电磁炉直火加热\"]}}",
      "modelKey": "custom",
      "versionLabel": "workspace"
    },
    {
      "id": "b",
      "label": "B",
      "testCaseId": "tc-1",
      "promptRef": {
        "kind": "version",
        "version": 4,
        "label": "Previous"
      },
      "promptText": "你是电商商品信息抽取助手。\n只输出 JSON 对象，字段必须为 title, selling_points, cautions。\n不要改字段名，不要添加外层包裹对象，不要解释。",
      "output": "{\"title\":\"便携手冲咖啡壶\",\"selling_points\":[\"600ml 容量\",\"适合露营和办公室\",\"双层不锈钢保温\"],\"cautions\":[\"不支持电磁炉直火加热\"]}",
      "modelKey": "custom",
      "versionLabel": "v4"
    },
    {
      "id": "c",
      "label": "C",
      "testCaseId": "tc-1",
      "promptRef": {
        "kind": "workspace",
        "label": "Teacher Workspace"
      },
      "promptText": "你是电商商品信息抽取助手。\n输出一个 JSON 对象，字段改为 product_name, buyer_highlights, cautions。\n为了后续扩展，把三个字段统一包在 payload 里。\n不要解释。",
      "output": "{\"payload\":{\"product_name\":\"便携手冲咖啡壶\",\"buyer_highlights\":[\"双场景使用：露营与办公室\",\"600ml 大容量\",\"双层不锈钢保温更稳\"],\"cautions\":[\"不支持电磁炉直火加热\"]}}",
      "modelKey": "deepseek",
      "versionLabel": "teacher-workspace"
    },
    {
      "id": "d",
      "label": "D",
      "testCaseId": "tc-1",
      "promptRef": {
        "kind": "version",
        "version": 4,
        "label": "Teacher Previous"
      },
      "promptText": "你是电商商品信息抽取助手。\n只输出 JSON 对象，字段必须为 title, selling_points, cautions。\n不要改字段名，不要添加外层包裹对象，不要解释。",
      "output": "{\"title\":\"便携手冲咖啡壶\",\"selling_points\":[\"600ml 容量\",\"适合露营和办公室\",\"双层不锈钢保温\"],\"cautions\":[\"不支持电磁炉直火加热\"]}",
      "modelKey": "deepseek",
      "versionLabel": "teacher-v4"
    }
  ],
  "compareHints": {
    "mode": "structured",
    "snapshotRoles": {
      "a": "target",
      "b": "baseline",
      "c": "reference",
      "d": "referenceBaseline"
    },
    "hasSharedTestCases": true,
    "hasSamePromptSnapshots": true,
    "hasCrossModelComparison": true
  }
}