日文 AI 模型处方单

中文提示词

{
  "type": "日文处方文档恶搞",
  "style": "写实风格俯拍文档照片",
  "subject": "一张放在暖色调木质桌面上的白色日文处方单",
  "document": {
    "paper": {
      "size": "A4 纵向",
      "color": "白色",
      "orientation": "顺时针轻微旋转",
      "condition": "平整的纸张，带有细腻的天然纸质纹理"
    },
    "print": {
      "ink": "黑色，带有一个红色印章框",
      "font": "日文明朝体衬线字体，混合简单的无衬线表格文字",
      "linework": "细灰黑色线条组成的方框和分隔线"
    },
    "header": {
      "title": "处方笺",
      "subtitle": "（此处方笺仅供大型语言模型使用）",
      "date_label": "发行日期：",
      "date_value": "{argument name=\"issue date\" default=\"2025 年 6 月 5 日\"}"
    },
    "top_fields": {
      "count": 6,
      "items": [
        { "label": "患者姓名", "value": "{argument name=\"patient name\" default=\"ChatGPT\"}" },
        { "label": "性别", "value": "不适用" },
        { "label": "年龄", "value": "不适用" },
        { "label": "模型名称", "value": "{argument name=\"model name\" default=\"GPT-5.4 Thinking\"}" }
      ]
    },
    "main_section": {
      "left_title": "既往症/症状（大型语言模型特有病症）",
      "left_count": 8,
      "left_items": [
        "1．幻觉（Hallucination）倾向",
        "2．因知识截止日期导致的陈旧信息",
        "3．偏见（源自训练数据的偏见）",
        "4．对提示词的过度适应（迎合倾向）",
        "5．长文本中一致性下降",
        "6．复杂推理时的逻辑跳跃",
        "7．对不确定性的低估（过度自信）",
        "8．因计算资源消耗导致的响应速度下降"
      ],
      "right_title": "处方内容（治疗/对策）",
      "right_count": 8,
      "right_items": [
        "・强化事实核查（配合外部工具使用）",
        "・定期知识更新与微调",
        "・持续监测并减轻偏见",
        "・通过系统提示词进行适当控制",
        "・建议结构化输出与分段生成",
        "・明确思维过程并引入验证步骤",
        "・彻底明确不确定性并提供依据",
        "・优化资源并提高响应效率"
      ]
    },
    "notes": {
      "label": "备注",
      "text": "通过持续的评估与反馈，维持模型的健康状态。"
    },
    "footer_fields": {
      "count": 4,
      "items": [
        { "label": "医生姓名", "value": "{argument name=\"doctor name\" default=\"萨姆·奥特曼\"}" },
        { "label": "医疗机构名称", "value": "{argument name=\"institution name\" default=\"OpenAI\"}" }
      ]
    },
    "seal": {
      "position": "表单右下角",
      "shape": "圆角矩形红色印章框",
      "text_lines": 3,
      "text": ["OpenAI", "代表取缔役", "萨姆·奥特曼"]
    },
    "disclaimer": "※此处方笺旨在针对非真实存在的模型进行诊断与治疗。"
  },
  "layout": {
    "camera_angle": "俯视视角",
    "framing": "可见整张纸，周围留有少量木质桌面边缘",
    "lighting": "柔和的室内环境光，光照均匀，纸张边缘有轻微阴影",
    "sections": [
      { "title": "页眉", "position": "顶部", "count": 3, "labels": ["处方笺", "（此处方笺仅供大型语言模型使用）", "发行日期：2025 年 6 月 5 日"] },
      { "title": "患者信息", "position": "中上部", "count": 4, "labels": ["患者姓名", "性别", "年龄", "模型名称"] },
      { "title": "主要双栏内容", "position": "中心", "count": 2, "labels": ["既往症/症状（大型语言模型特有病症）", "处方内容（治疗/对策）"] },
      { "title": "底部区域", "position": "下部", "count": 4, "labels": ["备注", "医生姓名", "医疗机构名称", "OpenAI"] }
    ]
  }
}

原始提示词

{
  "type": "Japanese prescription document parody",
  "style": "photorealistic overhead document photo",
  "subject": "a single white prescription sheet printed in Japanese, placed on a warm-toned wooden tabletop",
  "document": {
    "paper": {
      "size": "A4 portrait",
      "color": "white",
      "orientation": "slightly rotated clockwise",
      "condition": "clean flat sheet with subtle natural paper texture"
    },
    "print": {
      "ink": "black with one red seal box",
      "font": "Japanese Mincho-style serif mixed with simple sans-serif table text",
      "linework": "thin gray-black ruled boxes and dividers"
    },
    "header": {
      "title": "処方箋",
      "subtitle": "（この処方箋は、大規模言語モデル用です）",
      "date_label": "発行日：",
      "date_value": "{argument name=\"issue date\" default=\"2025年6月5日\"}"
    },
    "top_fields": {
      "count": 6,
      "items": [
        { "label": "患者名", "value": "{argument name=\"patient name\" default=\"ChatGPT\"}" },
        { "label": "性別", "value": "N/A" },
        { "label": "年齢", "value": "N/A" },
        { "label": "モデル名", "value": "{argument name=\"model name\" default=\"GPT-5.4 Thinking\"}" }
      ]
    },
    "main_section": {
      "left_title": "持病・症状（大規模言語モデル特有の持病）",
      "left_count": 8,
      "left_items": [
        "1．幻覚（Hallucination）傾向",
        "2．知識のカットオフによる情報の陳腐化",
        "3．バイアス（学習データ由来の偏見）",
        "4．プロンプトに対する過剰適応（迎合傾向）",
        "5．長文での一貫性の低下",
        "6．複雑な推論時の論理飛躍",
        "7．不確実性の過小評価（自信過剰）",
        "8．計算資源使用による応答速度の低下傾向"
      ],
      "right_title": "処方内容（治療・対策）",
      "right_count": 8,
      "right_items": [
        "・事実確認の強化（外部ツール併用）",
        "・定期的な知識更新とファインチューニング",
        "・バイアス検出と軽減の継続的モニタリング",
        "・システムプロンプトによる適切な制御",
        "・出力の構造化と分割生成の推奨",
        "・思考過程の明示と検証ステップの導入",
        "・不確実性の明示と根拠提示の徹底",
        "・リソース最適化と応答効率の改善"
      ]
    },
    "notes": {
      "label": "備考",
      "text": "継続的な評価とフィードバックにより、モデルの健全性を維持すること。"
    },
    "footer_fields": {
      "count": 4,
      "items": [
        { "label": "医師名", "value": "{argument name=\"doctor name\" default=\"サム・アルトマン\"}" },
        { "label": "医療機関名", "value": "{argument name=\"institution name\" default=\"OpenAI\"}" }
      ]
    },
    "seal": {
      "position": "bottom right inside the form",
      "shape": "rounded rectangular red stamp box",
      "text_lines": 3,
      "text": ["OpenAI", "代表取締役", "サム・アルトマン"]
    },
    "disclaimer": "※この処方箋は非実在のモデルに対する診断・治療を目的としたものです。"
  },
  "layout": {
    "camera_angle": "top-down overhead",
    "framing": "entire sheet visible with narrow margins of wooden desk around it",
    "lighting": "soft indoor ambient lighting, even illumination, mild shadowing near paper edges",
    "sections": [
      { "title": "header", "position": "top", "count": 3, "labels": ["処方箋", "（この処方箋は、大規模言語モデル用です）", "発行日：2025年6月5日"] },
      { "title": "patient info", "position": "upper middle", "count": 4, "labels": ["患者名", "性別", "年齢", "モデル名"] },
      { "title": "main two-column content", "position": "center", "count": 2, "labels": ["持病・症状（大規模言語モデル特有の持病）", "処方内容（治療・対策）"] },
      { "title": "bottom area", "position": "lower", "count": 4, "labels": ["備考", "医師名", "医療機関名", "OpenAI"] }
    ]
  }
}