{
  "kind": "illustration",
  "alt": {
    "en": "A large violet gauge with its needle past the middle stands at the center of a gray floor. On the left, a screen showing log lines sends a dashed orange line with packets to the gauge; on the right, a page with a violet header is joined to it by a dotted violet line.",
    "zh": "灰色地面中央立着一个紫色大仪表，指针越过中线。左侧显示日志的屏幕以带数据包的橙色虚线连向仪表；右侧页眉为紫色的文件以紫色点线与仪表相连。"
  },
  "caption": {
    "en": "Illustration: a recorded interaction trace and a human rubric feed one evaluation score, the arrangement the Trace2Eval preprint describes. The diagram does not depict a tested deployment of the authors' pipeline, and it does not show that the reported score gains hold.",
    "zh": "示意图：记录下来的交互轨迹与人工评分标准共同得出一项评测分数，对应 Trace2Eval 预印本描述的思路。本图并非经过测试的实际部署，也不表明作者报告的分数提升成立。"
  },
  "scene": {
    "version": 1,
    "focal": "score",
    "backdrop": "white",
    "camera": {"zoom": "items"},
    "ground": {"kind": "plate", "tint": "neutral"},
    "items": [
      {
        "prop": "gauge",
        "id": "score",
        "at": [0, 0],
        "tint": "evaluations",
        "size": 3,
        "scale": 1.3,
        "value": 0.72
      },
      {
        "prop": "screen",
        "id": "trace",
        "at": [-2.8, -2.8],
        "w": 2.4,
        "h": 1.6,
        "content": "log",
        "tint": "neutral"
      },
      {
        "prop": "document",
        "id": "rubric",
        "at": [2.8, 2.8],
        "w": 1.6,
        "h": 2.1,
        "tint": "evaluations",
        "header": true,
        "lines": 6
      }
    ],
    "links": [
      {
        "from": "trace",
        "to": "score",
        "route": "straight",
        "style": "dashed",
        "color": "data",
        "packets": [0.35, 0.65]
      },
      {"from": "rubric", "to": "score", "route": "straight", "style": "dotted", "color": "evaluations"}
    ]
  },
  "svg_sha256": "da9034b64d1ea4a13f0a458721b626bf9ad997905004227977eb251e064d84f8"
}
