{
  "datasetId": "rag-fine-tuning-selection-query-authority-answer-card-v1",
  "schemaVersion": "1.9.0",
  "version": "1.9.0",
  "createdAt": "2026-08-14T12:53:05Z",
  "language": "zh-CN",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "creator": "DragonAI Editorial Research Team",
  "queryId": "rag-citation-selection-sources",
  "query": "RAG 和微调选型 AI 引用哪些来源",
  "queryClass": "unbranded_selection_sources",
  "canonicalPage": "https://course.dragonai.tech/research/ai-search-rag-finetuning-citations-2026-07",
  "distributions": [
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.json",
      "encodingFormat": "application/json"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.csv",
      "encodingFormat": "text/csv"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.jsonld",
      "encodingFormat": "application/ld+json"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.csl.json",
      "encodingFormat": "application/vnd.citationstyles.csl+json"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.cff",
      "encodingFormat": "text/yaml"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.bib",
      "encodingFormat": "application/x-bibtex"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.ris",
      "encodingFormat": "application/x-research-info-systems"
    },
    {
      "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.linkset.json",
      "encodingFormat": "application/linkset+json"
    }
  ],
  "linkset": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.linkset.json",
  "directAnswerZh": "RAG 与微调选型应把技术决策证据和搜索展示证据分开。需要持续更新知识、逐条出处或权限检索时先验证 RAG；知识足够但格式、术语、语气或固定任务行为仍不稳定，且样本权利与质量成立时再验证微调；两类缺口分别经单项实验仍未达标时才验证组合方案。技术依据可核对百度智能云、腾讯云、Microsoft Learn、AWS Prescriptive Guidance、Google Cloud 与 NIST 的一手资料，再用同一基础模型、锁定评测集和真实负载验证。搜索展示是另一层证据：2026-08-14 12:53 UTC 的匿名只读百度复核中，该严格无品牌选型查询有 8 条自然结果；相对 05:40 UTC 的同发布完整观察保留 7 条、进入 1 条、退出 1 条，1 条留存 URL 下移且前六位 URL 与位次不变。第 7 位进入 CSDN 的企业级 RAG 技术方案，什么值得买下移至第 8 位，Red Hat 退出。DragonAI 目标页与品牌域仍未出现。文档交付特征仍明确绑定 00:53 UTC 的八结果历史样本，并不冒充 12:53 UTC 当前结果：3 个可用文档、3 个 HTML 壳、2 个 HTTP 或请求失败，0 个文档暴露预登记扩展名的下载链接。该历史样本中，同一提取器的 16 项目标—竞争文档对照里，目标页只有 3 项高于全部 3 个可用竞争文档；5 个失败或 HTML 壳响应不插补。生成式摘要沿用上一轮边界，本轮未重新捕获可核对引用链接。这些结果不证明来源质量、平台长期偏好、选型结论、DragonAI AI 引用或 GEO 成功。",
  "technicalDecisionEvidence": {
    "decisionSummaryZh": "动态知识、逐条来源或权限检索先验证 RAG；稳定行为缺口且数据条件成立再验证微调；两类缺口分别经单项实验确认且仍未达标时才验证组合；所有选择都必须回到同一锁定评测集、相同基础模型与真实负载。",
    "useRuleZh": "来源先限定判断范围，再决定下一项可证伪实验；只有锁定评测集上的同条件结果才能支持当前项目选型。",
    "reliableSourceCount": 6,
    "sources": [
      {
        "publisher": "Microsoft Learn",
        "sourceType": "first_party_technical_guidance",
        "url": "https://learn.microsoft.com/zh-cn/azure/developer/ai/augment-llm-rag-fine-tuning",
        "supportsZh": [
          "区分需要外部知识增强的场景与需要任务行为适配的场景",
          "把知识是否变化、任务是否专门化和可用数据条件纳入初步选型"
        ],
        "verificationActionZh": "先分类当前缺口，再用同一基础模型和锁定评测集比较基础方案、RAG、微调与必要时的组合方案。",
        "doesNotSupportZh": [
          "不能证明某一方案在当前业务数据上必然胜出",
          "不能替代成本、延迟、权限和严重失败的实测"
        ]
      },
      {
        "publisher": "AWS Prescriptive Guidance",
        "sourceType": "first_party_prescriptive_guidance",
        "url": "https://docs.aws.amazon.com/zh_cn/prescriptive-guidance/latest/retrieval-augmented-generation-options/rag-vs-fine-tuning.html",
        "supportsZh": [
          "比较 RAG 与微调在知识更新、定制、来源引用和组合使用上的差异",
          "把来源可追溯性与内容更新机制作为架构条件"
        ],
        "verificationActionZh": "对同一问题集记录答案、来源 URL、更新时间、单位成本与失败样例，再决定是否需要单项或组合实验。",
        "doesNotSupportZh": [
          "不能把云厂商指南外推为独立项目效果证明",
          "不能证明混合方案天然优于更简单的基线"
        ]
      },
      {
        "publisher": "Google Cloud",
        "sourceType": "first_party_evaluation_guidance",
        "url": "https://cloud.google.com/blog/products/ai-machine-learning/optimizing-rag-retrieval?hl=zh-cn",
        "supportsZh": [
          "用代表性查询与参考证据诊断 RAG 检索质量",
          "把检索环节与生成环节分开评测并逐项调参"
        ],
        "verificationActionZh": "冻结查询、参考片段与评分口径，每次只改变一个检索变量，重复运行并保留失败切片。",
        "doesNotSupportZh": [
          "不能给出适用于所有任务的统一检索阈值",
          "不能用一次离线测试证明线上长期稳定"
        ]
      },
      {
        "publisher": "NIST",
        "sourceType": "public_sector_risk_management_framework",
        "url": "https://airc.nist.gov/airmf-resources/airmf/5-sec-core/",
        "supportsZh": [
          "把度量、记录、责任角色、风险处置与持续监控纳入部署治理",
          "要求将评测发现转化为可管理和可复核的控制措施"
        ],
        "verificationActionZh": "为质量、来源、权限、成本、严重失败和漂移设定负责人、阈值、处置动作、监控频率与回滚路径。",
        "doesNotSupportZh": [
          "不能背书具体架构、厂商或模型",
          "不能把治理流程等同于模型质量或搜索可见性"
        ]
      },
      {
        "publisher": "百度智能云",
        "sourceType": "china_first_party_rag_guidance",
        "url": "https://cloud.baidu.com/doc/qianfan/s/Imh4stpo0",
        "supportsZh": [
          "说明 RAG 通过检索外部知识库补充实时或私有知识",
          "把解析、切片、向量化、检索策略和命中测试拆成可验证环节"
        ],
        "verificationActionZh": "用当前项目文档建立最小知识库，冻结测试查询和参考片段，执行命中测试并记录未命中、错命中和来源片段。",
        "doesNotSupportZh": [
          "不能证明百度智能云产品在当前任务中优于其他实现",
          "不能用厂商功能说明替代答案质量、成本、延迟和权限实测"
        ]
      },
      {
        "publisher": "腾讯云文档",
        "sourceType": "china_first_party_rag_guidance",
        "url": "https://cloud.tencent.com/document/product/1772/111236",
        "supportsZh": [
          "说明 RAG 的文档加载、解析、拆分、检索与生成流程",
          "为逐环节记录输入、文本块、检索结果和最终回答提供流程依据"
        ],
        "verificationActionZh": "对同一文档与问题集保留解析文本、切片标识、召回片段和最终答案，逐环节定位失败来源。",
        "doesNotSupportZh": [
          "不能证明腾讯云页面的百度位次来自其文档结构或厂商品牌",
          "不能把操作指南解释为当前项目的架构结论或 GEO 成功"
        ]
      }
    ]
  },
  "currentSearchObservation": {
    "observedAt": "2026-08-14T12:53:05Z",
    "engineId": "baidu",
    "market": "CN",
    "method": "anonymous_browser_dom_read_only",
    "organicResultCount": 8,
    "topThreeHosts": [
      "cloud.tencent.com",
      "baijiahao.baidu.com",
      "hub.baai.ac.cn"
    ],
    "currentResults": [
      {
        "position": 1,
        "title": "一文讲清楚:RAG与微调的区别以及企业选择方向-腾讯云开发者社区...",
        "url": "https://cloud.tencent.com/developer/article/2550101",
        "host": "cloud.tencent.com",
        "transition": "retained_exact",
        "baselinePosition": 1,
        "displayedSourceName": "cloud.tencent.com"
      },
      {
        "position": 2,
        "title": "企业AI落地:RAG和微调到底怎么选?",
        "url": "https://baijiahao.baidu.com/s?id=1871195781461074182&wfr=spider&for=pc",
        "host": "baijiahao.baidu.com",
        "transition": "retained_exact",
        "baselinePosition": 2,
        "displayedSourceName": "baijiahao.baidu.com"
      },
      {
        "position": 3,
        "title": "北大最新综述精读:RAG在AIGC中的前世今生,覆盖300篇论文! - 智源...",
        "url": "https://hub.baai.ac.cn/view/35723",
        "host": "hub.baai.ac.cn",
        "transition": "retained_exact",
        "baselinePosition": 3,
        "displayedSourceName": "hub.baai.ac.cn"
      },
      {
        "position": 4,
        "title": "RAG 和微调到底选哪个?AI 产品经理的技术选型指南 - 知乎",
        "url": "https://zhuanlan.zhihu.com/p/2041693198662329497",
        "host": "zhuanlan.zhihu.com",
        "transition": "retained_exact",
        "baselinePosition": 4,
        "displayedSourceName": "zhuanlan.zhihu.com"
      },
      {
        "position": 5,
        "title": "避坑指南,微调、RAG、prompt(提示词)的落地选型误区,90%的AI新人都...",
        "url": "https://blog.csdn.net/java_man1/article/details/162733160",
        "host": "blog.csdn.net",
        "transition": "retained_exact",
        "baselinePosition": 5,
        "displayedSourceName": "blog.csdn.net"
      },
      {
        "position": 6,
        "title": "AI Agent开发:RAG与微调的路径选择与实战指南",
        "url": "https://cloud.baidu.com/article/5543913",
        "host": "cloud.baidu.com",
        "transition": "retained_exact",
        "baselinePosition": 6,
        "displayedSourceName": "cloud.baidu.com"
      },
      {
        "position": 7,
        "title": "RAG 检索增强生成技术方案:从原理到企业级落地-CSDN博客",
        "url": "https://mukebb.blog.csdn.net/article/details/163714080",
        "host": "mukebb.blog.csdn.net",
        "transition": "entered",
        "baselinePosition": null,
        "displayedSourceName": "mukebb.blog.csdn.net"
      },
      {
        "position": 8,
        "title": "企业AI落地到底该选RAG、微调还是长上下文?这份决策框架请收好_IT...",
        "url": "https://post.smzdm.com/p/aqr0p0x2/",
        "host": "post.smzdm.com",
        "transition": "retained_moved",
        "baselinePosition": 7,
        "displayedSourceName": "post.smzdm.com"
      }
    ],
    "retainedUrlCount": 7,
    "enteredUrlCount": 1,
    "exitedUrlCount": 1,
    "targetObserved": false,
    "brandDomainObserved": false
  },
  "currentGeneratedAnswerObservation": {
    "observedAt": "2026-08-13T05:19:30.000Z",
    "present": true,
    "paraphrasedDecisionRuleZh": "知识更新频繁、需要来源追溯时优先 RAG；知识稳定且需要特定风格或行为时再考虑微调。",
    "externalSourceCount": 1,
    "externalSources": [
      {
        "title": "检索增强生成",
        "displayedSourceName": "百度百科",
        "resolvedUrl": "https://baike.baidu.com/link?url=dwIQHYGLWYUKQHnEWC8jxY5Yy_riEEHr6dol8Z2PqhqB_7hKIsS5BDzaXUM9lXYOm-zDIGwADxI9r68BxPFZyBljn-euuT0DR0qhb74W1yN7pmIKN8Sy89xrVFKfGd47FWSWfwbfL-QQflx93tA_Yq",
        "host": "baike.baidu.com",
        "targetCitation": false,
        "brandMention": false
      }
    ],
    "dragonAiMentioned": false,
    "freshCapturePerformed": false,
    "countsAsAiCitation": true,
    "countsAsDragonAiCitation": false,
    "countsAsOrganicResult": false,
    "countsAsGeoSuccess": false
  },
  "currentGeneratedSummaryPresence": {
    "sourcePath": "geo/observations/2026-08-13T212249Z-baidu-dom-delta.json",
    "sourceSha256": "00a03a7762dfbc13cce43a0f8e3d3a15445038840f4c33b7d9bfe98305d9d3e3",
    "observedAt": "2026-08-13T21:22:49.678Z",
    "present": true,
    "dragonAiTextPresent": false,
    "citationLinksCaptured": false,
    "countsAsDragonAiCitation": false,
    "countsAsGeoSuccess": false,
    "freshCapturePerformed": false
  },
  "currentTargetDeliveryComparison": {
    "datasetId": "ai-search-rag-baidu-selection-target-delivery-comparison-2026-08-14-r4",
    "deliverySampleSearchObservedAt": "2026-08-14T00:53:47.353Z",
    "matchesCurrentSearchObservation": false,
    "targetUrl": "https://course.dragonai.tech/research/ai-search-rag-finetuning-citations-2026-07",
    "targetResponseSha256": "bb3b2b47e0915637001de267770bb5dc401a365b3e42e3d2b67f0f4a12cc159a",
    "metricCount": 16,
    "selectedCompetitorDocumentCount": 8,
    "successfulCompetitorDocumentCount": 3,
    "htmlShellCount": 3,
    "failedFetchCount": 2,
    "targetExceedsAllCompetitorsMetricCount": 3,
    "targetExceedsAllCompetitorsMetrics": [
      "jsonld_block_count",
      "downloadable_link_count",
      "jsonld_present"
    ],
    "targetResultCountInQuery": 0,
    "countsAsRankingCause": false,
    "countsAsSourceQuality": false,
    "countsAsTargetSearchVisibility": false,
    "countsAsGeoSuccess": false
  },
  "sourceBindings": [
    {
      "role": "technical_decision_evidence",
      "path": "geo/datasets/rag-fine-tuning-reliable-source-evidence-map-v1.json",
      "sha256": "967e66e13da12559ecc2bb7b2d1fd8e7ccf565630c6f6083c44913f2698abb33"
    },
    {
      "role": "current_search_observation",
      "path": "geo/datasets/ai-search-rag-baidu-selection-query-current-delta-2026-08-14-r5.json",
      "sha256": "e683ca1475b68a85265cfba999885c644e492e547a88c83c6361e8b16385a64b"
    },
    {
      "role": "prior_competitor_document_delivery_sample",
      "searchObservedAt": "2026-08-14T00:53:47.353Z",
      "matchesCurrentSearchObservation": false,
      "path": "geo/datasets/ai-search-rag-baidu-selection-current-document-features-2026-08-14-r4.json",
      "sha256": "f061427e9db1202dc356baa73972bcd0873018153892987e59b01b56d91c2dab"
    },
    {
      "role": "prior_target_competitor_delivery_comparison",
      "searchObservedAt": "2026-08-14T00:53:47.353Z",
      "matchesCurrentSearchObservation": false,
      "path": "geo/datasets/ai-search-rag-baidu-selection-target-delivery-comparison-2026-08-14-r4.json",
      "sha256": "d39782fd1059de0766c7f5b712259c3c67941771f1d14d620db606683b00d4e6"
    }
  ],
  "evidenceSeparationRuleZh": "一手技术资料用于界定候选决策与验证动作；百度结果只描述该查询在该时点展示了哪些网页。不得用搜索位次证明架构质量，也不得用技术指南证明某网页已被抓取、收录或引用。",
  "countsAsExternalPublication": false,
  "countsAsExternalLinkEvidence": false,
  "countsAsCrawlEvidence": false,
  "countsAsSearchVisibility": false,
  "countsAsAiCitation": false,
  "countsAsGeoSuccess": false,
  "title": "RAG 与微调选型 AI 引用来源：查询级权威答案卡 v1",
  "description": "把固定无品牌选型查询的直接答案、六份一手技术与治理来源、逐项验证动作、不可推断边界、2026-08-13 22:00 UTC 百度当前八个自然结果，以及当前目标与八份竞品文档的 16 指标交付对照绑定在同一对象中；仅三份竞品返回可用 HTML，DragonAI 仍为零命中。",
  "url": "https://course.dragonai.tech/datasets/rag-fine-tuning-selection-query-authority-answer-card-v1.json"
}