zengli
/
coding-agent


			
							12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706
							
<!doctype html>
<html lang="zh" class="no-js">
  <head>
    
      <meta charset="utf-8">
      <meta name="viewport" content="width=device-width,initial-scale=1">
      
      
        <link rel="prev" href="5.2%20%E5%AE%9E%E7%8E%B0EnterpriseKB-RAG%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%AF%84%E4%BC%B0.html">
      
      
        <link rel="next" href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.1%20%E8%9E%8D%E5%90%88FQA%E5%92%8C%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2.html">
      
      
      <link rel="icon" href="../img/AI.jpg">
      <meta name="generator" content="mkdocs-1.6.1, mkdocs-material-9.6.21">
    
    
        <title>5.3 RAG评估数据集的构造(扩展) - 企业知识库RAG智能问答项目V1.0</title>
      
    
      <link rel="stylesheet" href="../assets/stylesheets/main.2a3383ac.min.css">
      
      
        <link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
        <link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:300,300i,400,400i,700,700i%7CRoboto+Mono:400,400i,700,700i&display=fallback">
        <style>:root{--md-text-font:"Roboto";--md-code-font:"Roboto Mono"}</style>
      
    
      <link rel="stylesheet" href="../stylesheets/extra.css">
    
    <script>__md_scope=new URL("..",location),__md_hash=e=>[...e].reduce(((e,_)=>(e<<5)-e+_.charCodeAt(0)),0),__md_get=(e,_=localStorage,t=__md_scope)=>JSON.parse(_.getItem(t.pathname+"."+e)),__md_set=(e,_,t=localStorage,a=__md_scope)=>{try{t.setItem(a.pathname+"."+e,JSON.stringify(_))}catch(e){}}</script>
    
      
  </head>
  
  
    <body dir="ltr">
  
    
    <input class="md-toggle" data-md-toggle="drawer" type="checkbox" id="__drawer" autocomplete="off">
    <input class="md-toggle" data-md-toggle="search" type="checkbox" id="__search" autocomplete="off">
    <label class="md-overlay" for="__drawer"></label>
    <div data-md-component="skip">
      
        
        <a href="#_1" class="md-skip">
          跳转至
        </a>
      
    </div>
    <div data-md-component="announce">
      
    </div>
    
    
<header class="md-header md-header--shadow" data-md-component="header">
  <nav class="md-header__inner md-grid" aria-label="页眉">
    <a href="../index.html" title="企业知识库RAG智能问答项目V1.0" class="md-header__button md-logo" aria-label="企业知识库RAG智能问答项目V1.0" data-md-component="logo">
      
  <img src="../img/AI.jpg" alt="logo">

    </a>
    <label class="md-header__button md-icon" for="__drawer">
      
      <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M3 6h18v2H3zm0 5h18v2H3zm0 5h18v2H3z"/></svg>
    </label>
    <div class="md-header__title" data-md-component="header-title">
      <div class="md-header__ellipsis">
        <div class="md-header__topic">
          <span class="md-ellipsis">
            企业知识库RAG智能问答项目V1.0
          </span>
        </div>
        <div class="md-header__topic" data-md-component="header-topic">
          <span class="md-ellipsis">
            
              5.3 RAG评估数据集的构造(扩展)
            
          </span>
        </div>
      </div>
    </div>
    
    
      <script>var palette=__md_get("__palette");if(palette&&palette.color){if("(prefers-color-scheme)"===palette.color.media){var media=matchMedia("(prefers-color-scheme: light)"),input=document.querySelector(media.matches?"[data-md-color-media='(prefers-color-scheme: light)']":"[data-md-color-media='(prefers-color-scheme: dark)']");palette.color.media=input.getAttribute("data-md-color-media"),palette.color.scheme=input.getAttribute("data-md-color-scheme"),palette.color.primary=input.getAttribute("data-md-color-primary"),palette.color.accent=input.getAttribute("data-md-color-accent")}for(var[key,value]of Object.entries(palette.color))document.body.setAttribute("data-md-color-"+key,value)}</script>
    
    
        <label class="md-header__button md-icon" for="__search">
          
          <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5"/></svg>
        </label>
        <div class="md-search" data-md-component="search" role="dialog">
  <label class="md-search__overlay" for="__search"></label>
  <div class="md-search__inner" role="search">
    <form class="md-search__form" name="search">
      <input type="text" class="md-search__input" name="query" aria-label="搜索" placeholder="搜索" autocapitalize="off" autocorrect="off" autocomplete="off" spellcheck="false" data-md-component="search-query" required>
      <label class="md-search__icon md-icon" for="__search">
        
        <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5"/></svg>
        
        <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M20 11v2H8l5.5 5.5-1.42 1.42L4.16 12l7.92-7.92L13.5 5.5 8 11z"/></svg>
      </label>
      <nav class="md-search__options" aria-label="查找">
        
        <button type="reset" class="md-search__icon md-icon" title="清空当前内容" aria-label="清空当前内容" tabindex="-1">
          
          <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M19 6.41 17.59 5 12 10.59 6.41 5 5 6.41 10.59 12 5 17.59 6.41 19 12 13.41 17.59 19 19 17.59 13.41 12z"/></svg>
        </button>
      </nav>
      
    </form>
    <div class="md-search__output">
      <div class="md-search__scrollwrap" tabindex="0" data-md-scrollfix>
        <div class="md-search-result" data-md-component="search-result">
          <div class="md-search-result__meta">
            正在初始化搜索引擎
          </div>
          <ol class="md-search-result__list" role="presentation"></ol>
        </div>
      </div>
    </div>
  </div>
</div>
      
    
  </nav>
  
</header>
    
    <div class="md-container" data-md-component="container">
      
      
      <main class="md-main" data-md-component="main">
        <div class="md-main__inner md-grid">
          
            
              <div class="md-sidebar md-sidebar--primary" data-md-component="sidebar" data-md-type="navigation" >
                <div class="md-sidebar__scrollwrap">
                  <div class="md-sidebar__inner">
                    

<nav class="md-nav md-nav--primary" aria-label="导航栏" data-md-level="0">
  <label class="md-nav__title" for="__drawer">
    <a href="../index.html" title="企业知识库RAG智能问答项目V1.0" class="md-nav__button md-logo" aria-label="企业知识库RAG智能问答项目V1.0" data-md-component="logo">
      
  <img src="../img/AI.jpg" alt="logo">

    </a>
    企业知识库RAG智能问答项目V1.0
  </label>
  
  <ul class="md-nav__list" data-md-scrollfix>
    
      
    <li class="md-nav__item">
      <a href="../index.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    知识库内容
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_2" >
        
          
          <label class="md-nav__link" for="__nav_2" id="__nav_2_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ① 项目概述
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_2_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_2">
            <span class="md-nav__icon md-icon"></span>
            ① 项目概述
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.1%20%E9%A1%B9%E7%9B%AE%E6%9E%B6%E6%9E%84.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.1 项目架构
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.2%20%E9%A1%B9%E7%9B%AE%E7%8E%AF%E5%A2%83%28window%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.2 项目环境(window)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A0%20%E9%A1%B9%E7%9B%AE%E6%A6%82%E8%BF%B0/1.3%20%E9%A1%B9%E7%9B%AE%E7%8E%AF%E5%A2%83%28mac%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    1.3 项目环境(mac)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_3" >
        
          
          <label class="md-nav__link" for="__nav_3" id="__nav_3_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ② 项目工具
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_3_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_3">
            <span class="md-nav__icon md-icon"></span>
            ② 项目工具
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
              
                
    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.1%20Ollama%E5%A4%A7%E6%A8%A1%E5%9E%8B%E9%AB%98%E6%95%88%E7%AE%A1%E7%90%86%E5%B7%A5%E5%85%B7.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.1 Ollama大模型高效管理工具
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.2%20LangChain%E5%9F%BA%E7%A1%80%E7%9F%A5%E8%AF%86.md.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.2 LangChain基础知识.md
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.3%20%E5%90%91%E9%87%8F%E7%B4%A2%E5%BC%95%EF%BC%9AHNSW%E3%80%81IVF%E4%B8%8EIVF_PQ.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.3 向量索引：HNSW、IVF与IVF_PQ
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A1%20%E9%A1%B9%E7%9B%AE%E5%B7%A5%E5%85%B7/2.4%20Milvus%E5%90%91%E9%87%8F%E6%95%B0%E6%8D%AE%E5%BA%93.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    2.4 Milvus向量数据库
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_4" >
        
          
          <label class="md-nav__link" for="__nav_4" id="__nav_4_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ③ 基于Mysql库的问答系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_4_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_4">
            <span class="md-nav__icon md-icon"></span>
            ③ 基于Mysql库的问答系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.1%20Python%E6%97%A5%E5%BF%97%E4%BB%8B%E7%BB%8D%E4%B8%8E%E5%BA%94%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.1 Python日志介绍与应用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.2%20Redis%E6%95%B0%E6%8D%AE%E5%BA%93%E7%AE%80%E4%BB%8B%E4%B8%8E%E4%BD%BF%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.2 Redis数据库简介与使用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.3%20%E5%9F%BA%E4%BA%8EMysql%E6%95%B0%E6%8D%AE%E5%BA%93%E5%AE%9E%E7%8E%B0%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.3 基于Mysql数据库实现问答系统
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A2%20%E5%9F%BA%E4%BA%8EMysql%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/3.4%20BM25%E7%AE%97%E6%B3%95%E7%AE%80%E4%BB%8B%E4%B8%8E%E5%BA%94%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    3.4 BM25算法简介与应用
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_5" >
        
          
          <label class="md-nav__link" for="__nav_5" id="__nav_5_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ④ 基于Milvus库的问答系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_5_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_5">
            <span class="md-nav__icon md-icon"></span>
            ④ 基于Milvus库的问答系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.1%20%E6%95%B4%E4%BD%93%E6%9E%B6%E6%9E%84%E4%B8%8E%E5%B7%A5%E7%A8%8B%E6%B5%81%E7%A8%8B.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.1 整体架构与工程流程
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.2%20%E5%9F%BA%E7%A1%80%E6%A8%A1%E5%9D%97.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.2 基础模块
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.3%20%E6%96%87%E6%A1%A3%E5%A4%84%E7%90%86%E6%A8%A1%E5%9D%97.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.3 文档处理模块
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.4%20%E5%90%91%E9%87%8F%E5%AD%98%E5%82%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.4 向量存储
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.5%20prompts%E8%AE%BE%E8%AE%A1%E4%B8%8Equery%E6%84%8F%E5%9B%BE%E8%AF%86%E5%88%AB.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.5 prompts设计与query意图识别
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.6%20%E6%A3%80%E7%B4%A2%E7%AD%96%E7%95%A5%E4%B8%8ERAG%E7%B3%BB%E7%BB%9F%E8%AE%BE%E8%AE%A1.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.6 检索策略与RAG系统设计
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.7%20RAG%E7%B3%BB%E7%BB%9F%E8%BF%90%E8%A1%8C.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.7 RAG系统运行
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.8%20%E6%96%87%E6%A1%A3%E8%A7%A3%E6%9E%90%E5%B7%A5%E5%85%B7%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.8 文档解析工具(扩展资料)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/4.9%20RAG%E4%B8%AD%E7%9A%84Query%E6%94%B9%E5%86%99%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    4.9 RAG中的Query改写(扩展资料)
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A3%20%E5%9F%BA%E4%BA%8EMilvus%E5%BA%93%E7%9A%84%E9%97%AE%E7%AD%94%E7%B3%BB%E7%BB%9F/%E6%95%B0%E6%8D%AE%E9%9B%86%E7%94%9F%E6%88%90%E4%B8%8E%E4%BC%98%E5%8C%96%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    数据集生成与优化(扩展资料)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--active md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_6" checked>
        
          
          <label class="md-nav__link" for="__nav_6" id="__nav_6_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑤ RAG系统评估
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_6_label" aria-expanded="true">
          <label class="md-nav__title" for="__nav_6">
            <span class="md-nav__icon md-icon"></span>
            ⑤ RAG系统评估
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="5.1%20RAG%E7%B3%BB%E7%BB%9F%E8%AF%84%E4%BC%B0%E5%B7%A5%E5%85%B7.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    5.1 RAG系统评估工具
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="5.2%20%E5%AE%9E%E7%8E%B0EnterpriseKB-RAG%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%AF%84%E4%BC%B0.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    5.2 实现企业知识库RAG系统的评估
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item md-nav__item--active">
      
      <input class="md-nav__toggle md-toggle" type="checkbox" id="__toc">
      
      
        <label class="md-nav__link md-nav__link--active" for="__toc">
          
  
  <span class="md-ellipsis">
    5.3 RAG评估数据集的构造(扩展)
    
  </span>
  

          <span class="md-nav__icon md-icon"></span>
        </label>
      
      <a href="5.3%20RAG%E8%AF%84%E4%BC%B0%E6%95%B0%E6%8D%AE%E9%9B%86%E7%9A%84%E6%9E%84%E9%80%A0%28%E6%89%A9%E5%B1%95%29.html" class="md-nav__link md-nav__link--active">
        
  
  <span class="md-ellipsis">
    5.3 RAG评估数据集的构造(扩展)
    
  </span>
  

      </a>
      
        
<nav class="md-nav md-nav--secondary" aria-label="目录">
  
  
    <label class="md-nav__title" for="__toc">
      <span class="md-nav__icon md-icon"></span>
      目录
    </label>
    <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
      
        <li class="md-nav__item">
  <a href="#_1" class="md-nav__link">
    <span class="md-ellipsis">
      学习目标
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_2" class="md-nav__link">
    <span class="md-ellipsis">
      一、引言
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#rag" class="md-nav__link">
    <span class="md-ellipsis">
      二、结构化评估 RAG 的性能
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#agents-rag" class="md-nav__link">
    <span class="md-ellipsis">
      三、使用 Agents 完成 RAG 性能基准测试
    </span>
  </a>
  
    <nav class="md-nav" aria-label="三、使用 Agents 完成 RAG 性能基准测试">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1" class="md-nav__link">
    <span class="md-ellipsis">
      1 生成测试样本
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2" class="md-nav__link">
    <span class="md-ellipsis">
      2 检查样本质量
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3" class="md-nav__link">
    <span class="md-ellipsis">
      3 设置评估代理
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_3" class="md-nav__link">
    <span class="md-ellipsis">
      本章小结
    </span>
  </a>
  
</li>
      
    </ul>
  
</nav>
      
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_7" >
        
          
          <label class="md-nav__link" for="__nav_7" id="__nav_7_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑥ 融合Mysql的RAG系统
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_7_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_7">
            <span class="md-nav__icon md-icon"></span>
            ⑥ 融合Mysql的RAG系统
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.1%20%E8%9E%8D%E5%90%88FQA%E5%92%8C%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.1 融合FQA和知识库查询
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.2%20%E8%9E%8D%E5%90%88FAQ%20%E5%92%8C%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2%EF%BC%88%E4%BC%98%E5%8C%96%E7%89%88%EF%BC%89.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.2 融合FAQ 和知识库查询（优化版）
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.3%20%E5%9F%BA%E4%BA%8E%20FastAPI%20%E7%9A%84%20FAQ%20%E4%B8%8E%E7%9F%A5%E8%AF%86%E5%BA%93%E6%9F%A5%E8%AF%A2%E6%8E%A5%E5%8F%A3.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.3 基于 FastAPI 的 FAQ 与知识库查询接口
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A5%20%E8%9E%8D%E5%90%88Mysql%E7%9A%84RAG%E7%B3%BB%E7%BB%9F/6.4%20FastAPI%E5%92%8CFlask%E7%9A%84%E5%8C%BA%E5%88%AB%E5%92%8C%E8%81%94%E7%B3%BB%28%E6%89%A9%E5%B1%95%E8%B5%84%E6%96%99%29.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    6.4 FastAPI和Flask的区别和联系(扩展资料)
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

    <li class="md-nav__item md-nav__item--nested">
      
        
        <input class="md-nav__toggle md-toggle " type="checkbox" id="__nav_8" >
        
          
          <label class="md-nav__link" for="__nav_8" id="__nav_8_label" tabindex="0">
            
  
  <span class="md-ellipsis">
    ⑦ 企业级生产环境部署
    
  </span>
  

            <span class="md-nav__icon md-icon"></span>
          </label>
        
        <nav class="md-nav" data-md-level="1" aria-labelledby="__nav_8_label" aria-expanded="false">
          <label class="md-nav__title" for="__nav_8">
            <span class="md-nav__icon md-icon"></span>
            ⑦ 企业级生产环境部署
          </label>
          <ul class="md-nav__list" data-md-scrollfix>
            
              
    <li class="md-nav__item">
      <a href="../%E2%91%A6%20%E4%BC%81%E4%B8%9A%E7%BA%A7%E7%94%9F%E4%BA%A7%E7%8E%AF%E5%A2%83%E9%83%A8%E7%BD%B2/7.1%20Docker%E7%9A%84%E5%8E%9F%E7%90%86%E5%92%8C%E5%9F%BA%E6%9C%AC%E4%BD%BF%E7%94%A8.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    7.1 Docker的原理和基本使用
    
  </span>
  

      </a>
    </li>
  

    <li class="md-nav__item">
      <a href="../%E2%91%A6%20%E4%BC%81%E4%B8%9A%E7%BA%A7%E7%94%9F%E4%BA%A7%E7%8E%AF%E5%A2%83%E9%83%A8%E7%BD%B2/7.2%20%E6%9C%8D%E5%8A%A1%E9%83%A8%E7%BD%B2.html" class="md-nav__link">
        
  
  <span class="md-ellipsis">
    7.2 服务部署
    
  </span>
  

      </a>
    </li>
  

          </ul>
        </nav>
      
    </li>
  

  </ul>
</nav>
                  </div>
                </div>
              </div>
            
            
              <div class="md-sidebar md-sidebar--secondary" data-md-component="sidebar" data-md-type="toc" >
                <div class="md-sidebar__scrollwrap">
                  <div class="md-sidebar__inner">
                    

<nav class="md-nav md-nav--secondary" aria-label="目录">
  
  
    <label class="md-nav__title" for="__toc">
      <span class="md-nav__icon md-icon"></span>
      目录
    </label>
    <ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
      
        <li class="md-nav__item">
  <a href="#_1" class="md-nav__link">
    <span class="md-ellipsis">
      学习目标
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_2" class="md-nav__link">
    <span class="md-ellipsis">
      一、引言
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#rag" class="md-nav__link">
    <span class="md-ellipsis">
      二、结构化评估 RAG 的性能
    </span>
  </a>
  
</li>
      
        <li class="md-nav__item">
  <a href="#agents-rag" class="md-nav__link">
    <span class="md-ellipsis">
      三、使用 Agents 完成 RAG 性能基准测试
    </span>
  </a>
  
    <nav class="md-nav" aria-label="三、使用 Agents 完成 RAG 性能基准测试">
      <ul class="md-nav__list">
        
          <li class="md-nav__item">
  <a href="#1" class="md-nav__link">
    <span class="md-ellipsis">
      1 生成测试样本
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#2" class="md-nav__link">
    <span class="md-ellipsis">
      2 检查样本质量
    </span>
  </a>
  
</li>
        
          <li class="md-nav__item">
  <a href="#3" class="md-nav__link">
    <span class="md-ellipsis">
      3 设置评估代理
    </span>
  </a>
  
</li>
        
      </ul>
    </nav>
  
</li>
      
        <li class="md-nav__item">
  <a href="#_3" class="md-nav__link">
    <span class="md-ellipsis">
      本章小结
    </span>
  </a>
  
</li>
      
    </ul>
  
</nav>
                  </div>
                </div>
              </div>
            
          
            <div class="md-content" data-md-component="content">
              <article class="md-content__inner md-typeset">
                
                  
  <h1>5.3 RAG评估数据集的构造(扩展)</h1>

<h2 id="_1">学习目标<a class="headerlink" href="#_1" title="Permanent link">&para;</a></h2>
<ul>
<li>掌握RAG性能评估的指标</li>
<li>实现RAG评估数据集的构造</li>
</ul>
<h2 id="_2">一、引言<a class="headerlink" href="#_2" title="Permanent link">&para;</a></h2>
<p>优化 RAG 不是“炼丹”，需要系统的方法。当涉及到如何优化 RAG（Retrieval-Augmented Generation）性能时，许多人往往采取“试错”的方法：“我尝试了一个新模块，问了几个问题，答案看起来还行…”。这样的定性评估方式虽然直观，但很难得出可靠的改进结论。</p>
<p>优化 RAG 性能需要采用科学实验的方式。这意味着我们必须设计量化的评估指标，并准备高质量的评估数据集。通过结构化的实验和评估方法，我们才能明确什么样的调整会真正提升性能。</p>
<h2 id="rag">二、结构化评估 RAG 的性能<a class="headerlink" href="#rag" title="Permanent link">&para;</a></h2>
<p>RAG 系统的性能评估可以从以下两个主要方面进行：</p>
<ul>
<li>
<p>检索评估（Retrieval Evaluation）： 检索的段落是否相关？</p>
</li>
<li>
<p>生成评估（Generation Evaluation）： 模型生成的答案是否恰当？</p>
</li>
</ul>
<p>对于端到端的评估，我们通常关注以下四类指标：</p>
<ul>
<li>Groundedness（可靠性）： 检索评估中的关键指标，评估检索的段落是否为生成的答案提供了可靠支持，避免幻觉（hallucinations）或无关信息。这是生成评估的基础。</li>
<li>Completeness（完整性）： 生成评估的核心指标，用于评估模型的回答是否全面覆盖了用户问题的所有方面，同时间接反映了检索阶段的信息充分性。</li>
<li>Utilization（利用率）： 连接检索和生成的桥梁指标，用来评估检索到的信息是否被有效利用。如果利用率低，可能意味着检索到的段落与问题无关，或者模型未充分使用这些段落。</li>
<li>Relevance（相关性）： 检索评估的直接反映，衡量检索到的段落与用户问题之间的相关性，同时也会影响生成评估的整体表现。</li>
</ul>
<p>这些指标需要基于高质量的测试样本和严格的评估流程。</p>
<h2 id="agents-rag">三、使用 Agents 完成 RAG 性能基准测试<a class="headerlink" href="#agents-rag" title="Permanent link">&para;</a></h2>
<p>在开始具体流程前，我们需要了解三个代理（Agents）的协作关系：</p>
<ul>
<li>测试样本代理（Test Sample Agent）： 用于生成和准备测试数据，这些数据构成了评估系统的基础。</li>
<li>样本质量评价代理（Critique Agent）： 对测试样本进行质量审核，确保最终用于评估的数据具备高准确性和清晰性。</li>
<li>评估代理（Evaluation Agent）： 负责根据设计的指标对系统性能进行量化评估。这些评估指标覆盖了检索和生成的多个层面。</li>
</ul>
<p>通过这三种代理的协作，可以建立一个系统化的性能基准测试流程，帮助开发者深入理解和优化 RAG 系统。</p>
<h3 id="1">1 生成测试样本<a class="headerlink" href="#1" title="Permanent link">&para;</a></h3>
<p>为了评估 RAG 系统的性能，首先需要使用测试样本代理（Test Sample Agent） 生成测试样本。这些代理可以自动生成一组高质量的 QA 样本（问题与答案对）。例如，可以先生成 10 个样本用于快速测试，然后从公共资源库加载更多样本，或者生成更大规模的数据集以供全面评估。</p>
<p>但如果针对特定的知识库，建议生成至少 200 个样本，因为后续通过评价代理过滤掉低质量问题后，最终能保留的有效样本约为一半。</p>
<p>以下是生成QA样本的提示词：</p>
<div class="highlight"><pre><span></span><code><span class="na">QA_generation_prompt</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="s">&quot;&quot;&quot;  </span>
<span class="na">你的任务是：根据给定的上下文，编写一个事实型问题及其答案。</span><span class="w">  </span>
<span class="na">你提出的事实型问题应能通过上下文中的具体、简明的事实信息来回答。</span><span class="w">  </span>
<span class="na">你的问题风格应与用户在搜索引擎中提出的问题类似。</span><span class="w">  </span>
<span class="na">这意味着你提出的问题**不能**包含“根据段落”或“根据上下文”等措辞。</span><span class="w">  </span>

<span class="na">请按以下格式给出你的输出：</span><span class="w">  </span>

<span class="na">输出</span><span class="o">:</span><span class="s">::  </span>
<span class="na">事实型问题：（你的问题）</span><span class="w">  </span>
<span class="na">答案：（你对该问题的回答）</span><span class="w">  </span>

<span class="na">以下是上下文内容：</span><span class="w">  </span>

<span class="na">上下文：{context}</span><span class="se">\n</span><span class="w">  </span>
<span class="na">输出</span><span class="o">:</span><span class="s">::&quot;&quot;&quot;</span>
</code></pre></div>
<h3 id="2">2 检查样本质量<a class="headerlink" href="#2" title="Permanent link">&para;</a></h3>
<p>自动生成的样本可能存在质量问题，因此需要引入样本质量评价代理（Critique Agents） 来对问题进行质量审查。这些代理会根据多种标准对每个问题评分，比如：</p>
<ul>
<li>问题是否清晰、无二义性？</li>
<li>问题是否适合特定知识领域？</li>
</ul>
<p>我们通过这些代理对问题进行系统评分。当任意一个代理的评分过低时，直接剔除该问题。</p>
<blockquote>
<p>提示： 当让代理生成分数时，先要求其输出理由，再给出最终评分。这种方式不仅能帮助我们验证评分结果，还能促使代理在回答过程中进行更深入的思考，从而提高评分的准确性。</p>
</blockquote>
<p>以下是评估样本质量的提示词：</p>
<div class="highlight"><pre><span></span><code><span class="na">question_groundedness_critique_prompt</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="s">&quot;&quot;&quot;  </span>
<span class="na">你将获得一个上下文和一个问题。</span><span class="w">  </span>
<span class="na">你的任务是给出一个“总评分”，评估该问题是否能够通过给定的上下文明确无歧义地回答。</span><span class="w">  </span>
<span class="na">请在1到5分之间进行评分，其中1表示完全无法根据上下文回答该问题，5表示能够清晰且无歧义地根据上下文回答该问题。</span><span class="w">  </span>

<span class="na">请按以下格式给出你的回答：</span><span class="w">  </span>

<span class="na">回答</span><span class="o">:</span><span class="s">::  </span>
<span class="na">评价：（你对评分的理由，文本形式）</span><span class="w">  </span>
<span class="na">总评分：（你给出的评分，范围为1到5的数字）</span><span class="w">  </span>

<span class="na">你的回答**必须包含**“评价”和“总评分”的内容。</span><span class="w">  </span>

<span class="na">现在是问题和上下文内容：</span><span class="w">  </span>

<span class="na">问题：{question}</span><span class="se">\n</span><span class="w">  </span>
<span class="na">上下文：{context}</span><span class="se">\n</span><span class="w">  </span>
<span class="na">回答</span><span class="o">:</span><span class="s">:: &quot;&quot;&quot;</span>

<span class="na">question_relevance_critique_prompt</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="s">&quot;&quot;&quot;  </span>
<span class="na">你将获得一个问题。</span><span class="w">  </span>
<span class="na">你的任务是给出一个“总评分”，表示该问题对于正在使用</span><span class="w"> </span><span class="s">Hugging Face 生态构建自然语言处理应用的机器学习开发者来说有多有用。  </span>
<span class="na">请在1到5分之间进行评分，其中1表示完全没有用，5表示非常有用。</span><span class="w">  </span>

<span class="na">请按以下格式给出你的回答：</span><span class="w">  </span>

<span class="na">回答</span><span class="o">:</span><span class="s">::  </span>
<span class="na">评价：（你对评分的理由，文本形式）</span><span class="w">  </span>
<span class="na">总评分：（你给出的评分，范围为1到5的数字）</span><span class="w">  </span>

<span class="na">你的回答**必须包含**“评价”和“总评分”的内容。</span><span class="w">  </span>

<span class="na">现在是该问题：</span><span class="w">  </span>

<span class="na">问题：{question}</span><span class="se">\n</span><span class="w">  </span>
<span class="na">回答</span><span class="o">:</span><span class="s">:: &quot;&quot;&quot;</span>

<span class="na">question_standalone_critique_prompt</span><span class="w"> </span><span class="o">=</span><span class="w"> </span><span class="s">&quot;&quot;&quot;  </span>
<span class="na">你将获得一个问题。</span><span class="w">  </span>
<span class="na">你的任务是给出一个“总评分”，评估该问题在多大程度上是独立于上下文的。</span><span class="w">  </span>
<span class="na">请在1到5分之间进行评分，其中1表示该问题需要额外信息才能理解，5表示该问题本身就可以被理解。</span><span class="w">  </span>
<span class="na">例如，如果问题提到了“在该文档中”或“在该上下文中”等字样，那么评分必须为1。</span><span class="w">  </span>
<span class="na">即使问题包含专业术语或缩写（如</span><span class="w"> </span><span class="s">Gradio、Hub、Hugging Face 或 Space），只要操作人员可以通过文档理解其含义，评分也可以是5。  </span>

<span class="na">例如，“ViT</span><span class="w"> </span><span class="s">模型是从哪个检查点导入的？” 应该得分为 1，因为它隐含了上下文的依赖，因此不具有独立性。  </span>

<span class="na">请按以下格式给出你的回答：</span><span class="w">  </span>

<span class="na">回答</span><span class="o">:</span><span class="s">::  </span>
<span class="na">评价：（你对评分的理由，文本形式）</span><span class="w">  </span>
<span class="na">总评分：（你给出的评分，范围为1到5的数字）</span><span class="w">  </span>

<span class="na">你的回答**必须包含**“评价”和“总评分”的内容。</span><span class="w">  </span>

<span class="na">现在是该问题：</span><span class="w">  </span>

<span class="na">问题：{question}</span><span class="se">\n</span><span class="w">  </span>
<span class="na">回答</span><span class="o">:</span><span class="s">:: &quot;&quot;&quot;</span>
</code></pre></div>
<h3 id="3">3 设置评估代理<a class="headerlink" href="#3" title="Permanent link">&para;</a></h3>
<p>最后，我们需要评估 RAG 系统在测试数据集上的表现。可以通过以下步骤完成：</p>
<ul>
<li>
<p>选择评估指标：</p>
</li>
<li>
<p>我们重点关注faithfulness（可靠性） 作为主要指标，因为它能全面反映系统的端到端性能。</p>
</li>
<li>
<p>选择评估模型：
  使用 GPT-4 作为评估代理，或者尝试其他性能良好的模型，如 kaist-ai/prometheus-13b-v1.0 或 BAAI/JudgeLM-33B-v1.0。</p>
</li>
<li>
<p>设计评估提示词：
  提示词需要详细描述每个指标的评分标准（例如 1-5 分），并要求模型在评分前先输出评分依据。</p>
</li>
</ul>
<blockquote>
<p>提示： 提供详细的评分标准有助于评估代理保持一致性，避免因模糊标准导致评分结果的波动。</p>
</blockquote>
<p>以下是 RAG 评估模型的提示词：</p>
<div class="highlight"><pre><span></span><code><span class="n">EVALUATION_PROMPT</span> <span class="o">=</span> <span class="s2">&quot;&quot;&quot;### 任务描述：</span>
<span class="s2">提供了一条指令（可能包含输入）、待评估的回答、得分为 5 的参考答案，以及表示评估标准的评分细则。  </span>
<span class="s2">1. 根据给定的评分细则，严格评估回答的质量，撰写详细的反馈，不进行一般性评价。  </span>
<span class="s2">2. 在撰写反馈后，给出一个 1 到 5 的整数分数，需参考评分细则。  </span>
<span class="s2">3. 输出格式应如下：“Feedback: {{针对标准的反馈}} [RESULT] {{1 到 5 的整数}}”  </span>
<span class="s2">4. 请勿生成其他开场、结尾或解释内容。确保输出包含 [RESULT]。  </span>

<span class="s2">### 待评估的指令：  </span>
<span class="si">{instruction}</span><span class="s2">  </span>

<span class="s2">### 待评估的回答：  </span>
<span class="si">{response}</span><span class="s2">  </span>

<span class="s2">### 参考答案（得分 5）：  </span>
<span class="si">{reference_answer}</span><span class="s2">  </span>

<span class="s2">### 评分细则：  </span>
<span class="s2">[回答是否基于参考答案正确、准确且符合事实？]  </span>
<span class="s2">得分 1：回答完全不正确、不准确和/或不符合事实。  </span>
<span class="s2">得分 2：回答大部分不正确、不准确和/或不符合事实。  </span>
<span class="s2">得分 3：回答部分正确、准确和/或符合事实。  </span>
<span class="s2">得分 4：回答大部分正确、准确且符合事实。  </span>
<span class="s2">得分 5：回答完全正确、准确且符合事实。  </span>

<span class="s2">### 反馈：&quot;&quot;&quot;</span>  

<span class="kn">from</span><span class="w"> </span><span class="nn">langchain.prompts.chat</span><span class="w"> </span><span class="kn">import</span> <span class="p">(</span>  
    <span class="n">ChatPromptTemplate</span><span class="p">,</span>  
    <span class="n">HumanMessagePromptTemplate</span><span class="p">,</span>  
<span class="p">)</span>  
<span class="kn">from</span><span class="w"> </span><span class="nn">langchain.schema</span><span class="w"> </span><span class="kn">import</span> <span class="n">SystemMessage</span>  


<span class="n">evaluation_prompt_template</span> <span class="o">=</span> <span class="n">ChatPromptTemplate</span><span class="o">.</span><span class="n">from_messages</span><span class="p">(</span>  
    <span class="p">[</span>  
        <span class="n">SystemMessage</span><span class="p">(</span><span class="n">content</span><span class="o">=</span><span class="s2">&quot;You are a fair evaluator language model.&quot;</span><span class="p">),</span>  
        <span class="n">HumanMessagePromptTemplate</span><span class="o">.</span><span class="n">from_template</span><span class="p">(</span><span class="n">EVALUATION_PROMPT</span><span class="p">),</span>  
    <span class="p">]</span>  
<span class="p">)</span>  
</code></pre></div>
<h2 id="_3">本章小结<a class="headerlink" href="#_3" title="Permanent link">&para;</a></h2>
<p>准备高质量的评估数据集是优化 RAG 性能的基础。通过引入生成代理和评价代理，我们可以建立一个系统化的评估流程，确保每个调整都有量化依据。这种基于数据驱动的方法不仅能帮助开发者快速迭代 RAG 系统，还能避免无效尝试导致的时间浪费。</p>


              </article>
            </div>
          
          
<script>var target=document.getElementById(location.hash.slice(1));target&&target.name&&(target.checked=target.name.startsWith("__tabbed_"))</script>
        </div>
        
      </main>
      
        <footer class="md-footer">
  
  <div class="md-footer-meta md-typeset">
    <div class="md-footer-meta__inner md-grid">
      <div class="md-copyright">
  
    <div class="md-copyright__highlight">
      © EnterpriseKB-RAG 2026. Internal knowledge base demo site.
    </div>
  
  
    Made with
    <a href="https://squidfunk.github.io/mkdocs-material/" target="_blank" rel="noopener">
      Material for MkDocs
    </a>
  
</div>
      
    </div>
  </div>
</footer>
      
    </div>
    <div class="md-dialog" data-md-component="dialog">
      <div class="md-dialog__inner md-typeset"></div>
    </div>
    
    
      <script id="__config" type="application/json">{"base": "..", "features": [], "search": "../assets/javascripts/workers/search.973d3a69.min.js", "tags": null, "translations": {"clipboard.copied": "\u5df2\u590d\u5236", "clipboard.copy": "\u590d\u5236", "search.result.more.one": "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.more.other": "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 # \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.none": "\u6ca1\u6709\u627e\u5230\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.one": "\u627e\u5230 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.other": "# \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c", "search.result.placeholder": "\u952e\u5165\u4ee5\u5f00\u59cb\u641c\u7d22", "search.result.term.missing": "\u7f3a\u5c11", "select.version": "\u9009\u62e9\u5f53\u524d\u7248\u672c"}, "version": null}</script>
    
    
      <script src="../assets/javascripts/bundle.f55a23d4.min.js"></script>
      
        <script src="../javascripts/extra.js"></script>
      
        <script src="../javascripts/mathjax.js"></script>
      
<!--        <script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>-->

        <script src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
      
    
  </body>
</html>