(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[16129],{84798:function(e,n,a){(window.__NEXT_P=window.__NEXT_P||[]).push(["/techniques/rag.fr",function(){return a(46598)}])},11196:function(e,n){"use strict";n.Z={src:"/_next/static/media/rag.c6528d99.png",height:492,width:960,blurDataURL:"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAgAAAAECAMAAACEE47CAAAAJ1BMVEXk5+Xg4+Dw8fDq+Pj1+PrR2+vm5vbK19W709jo3e/d2OjY19vl8e9DR0x8AAAACXBIWXMAAAsTAAALEwEAmpwYAAAAJUlEQVR4nAXBiQEAMAgCsQPUfvvv24RU2ZUwG6l78EKv7wEkwB8ISgBrVzHOiQAAAABJRU5ErkJggg==",blurWidth:8,blurHeight:4}},46598:function(e,n,a){"use strict";a.r(n),a.d(n,{__toc:function(){return u}});var t=a(11527),i=a(77154),o=a(13682),r=a(12540);a(13777),a(18525);var s=a(5424),l=a(68815),d=a(27728),c=a(94699),m=a(11196);let u=[{depth:2,value:"Cas d'utilisation d'un RAG : G\xe9n\xe9rer des titres d'articles de machine learning facilement.",id:"cas-dutilisation-dun-rag--g\xe9n\xe9rer-des-titres-darticles-de-machine-learning-facilement"},{depth:2,value:"R\xe9f\xe9rences",id:"r\xe9f\xe9rences"}];function _createMdxContent(e){let n=Object.assign({h1:"h1",p:"p",a:"a",h2:"h2",ul:"ul",li:"li"},(0,s.a)(),e.components);return(0,t.jsxs)(t.Fragment,{children:[(0,t.jsx)(n.h1,{children:"G\xe9n\xe9ration Augment\xe9e par R\xe9cup\xe9ration (RAG)"}),"\n",(0,t.jsx)(n.p,{children:'En anglais : "Retrieval Augmented Generation", commun\xe9ment abr\xe9g\xe9 en RAG'}),"\n","\n",(0,t.jsx)(n.p,{children:"Les mod\xe8les de langage polyvalents peuvent \xeatre affin\xe9s pour r\xe9aliser plusieurs t\xe2ches courantes telles que l'analyse de sentiments et la reconnaissance d'entit\xe9s nomm\xe9es. Ces t\xe2ches ne n\xe9cessitent g\xe9n\xe9ralement pas de connaissances suppl\xe9mentaires."}),"\n",(0,t.jsx)(n.p,{children:'Pour des t\xe2ches plus complexes et exigeantes en connaissances, il est possible de construire un syst\xe8me bas\xe9 sur un mod\xe8le de langage qui acc\xe8de \xe0 des sources de connaissances externes pour compl\xe9ter les t\xe2ches. Cela permet une plus grande coh\xe9rence factuelle, am\xe9liore la fiabilit\xe9 des r\xe9ponses g\xe9n\xe9r\xe9es et aide \xe0 att\xe9nuer le probl\xe8me des "hallucinations".'}),"\n",(0,t.jsx)(n.p,{children:"Les chercheurs de Meta AI ont introduit une m\xe9thode appel\xe9e G\xe9n\xe9ration Augment\xe9e par R\xe9cup\xe9ration (RAG) pour aborder de telles t\xe2ches exigeantes en connaissances. RAG combine un composant de r\xe9cup\xe9ration d'informations avec un mod\xe8le g\xe9n\xe9rateur de texte. RAG peut \xeatre affin\xe9 et ses connaissances internes peuvent \xeatre modifi\xe9es de mani\xe8re efficace et sans n\xe9cessiter une reformation compl\xe8te du mod\xe8le."}),"\n",(0,t.jsx)(n.p,{children:"RAG prend une entr\xe9e et r\xe9cup\xe8re un ensemble de documents pertinents/supportants donn\xe9s par une source (par exemple, Wikip\xe9dia). Les documents sont concat\xe9n\xe9s comme contexte avec la demande d'entr\xe9e originele et fournis au g\xe9n\xe9rateur de texte qui produit la sortie finale. Cela rend le RAG adaptable pour des situations o\xf9 les faits pourraient \xe9voluer avec le temps. Cela est tr\xe8s utile car la connaissance param\xe9trique des LLM est statique. le RAG permet aux mod\xe8les de langage de contourner la reformation, permettant l'acc\xe8s aux informations les plus r\xe9centes pour g\xe9n\xe9rer des sorties fiables via la g\xe9n\xe9ration bas\xe9e sur la r\xe9cup\xe9ration."}),"\n",(0,t.jsx)(n.p,{children:"Lewis et al., (2021) ont propos\xe9 une recette de raffinement polyvalente pour le RAG. Un mod\xe8le seq2seq pr\xe9-entra\xeen\xe9 est utilis\xe9 comme m\xe9moire param\xe9trique et un index vectoriel dense de Wikip\xe9dia est utilis\xe9 comme m\xe9moire non param\xe9trique (acc\xe9d\xe9e \xe0 l'aide d'un r\xe9cup\xe9rateur pr\xe9-entra\xeen\xe9 par r\xe9seau de neurones). Voici un aper\xe7u de la fa\xe7on dont l'approche fonctionne :"}),"\n",(0,t.jsx)(c.w,{src:m.Z,alt:"RAG"}),"\n",(0,t.jsxs)(n.p,{children:["Source de l'image : ",(0,t.jsx)(n.a,{href:"https://arxiv.org/pdf/2005.11401.pdf",children:"Lewis et el. (2021)"}),"\nUn RAG est performant sur plusieurs benchmarks tels que Natural Questions, WebQuestions, et CuratedTrec. Le RAG g\xe9n\xe8re des r\xe9ponses plus factuelles, sp\xe9cifiques et diversifi\xe9es lorsqu'il est test\xe9 sur des questions MS-MARCO et Jeopardy. le RAG am\xe9liore \xe9galement les r\xe9sultats sur la v\xe9rification des faits FEVER."]}),"\n",(0,t.jsx)(n.p,{children:"Cela montre le potentiel de RAG comme une option viable et pertinente pour am\xe9liorer les sorties des mod\xe8les de langage dans des t\xe2ches exigeantes en connaissances."}),"\n",(0,t.jsx)(n.p,{children:"Plus r\xe9cemment, ces approches bas\xe9es sur la r\xe9cup\xe9ration sont devenues plus populaires et sont combin\xe9es avec des LLM bien connus comme ChatGPT pour am\xe9liorer les capacit\xe9s et la coh\xe9rence factuelle."}),"\n",(0,t.jsx)(n.h2,{id:"cas-dutilisation-dun-rag--g\xe9n\xe9rer-des-titres-darticles-de-machine-learning-facilement",children:"Cas d'utilisation d'un RAG : G\xe9n\xe9rer des titres d'articles de machine learning facilement."}),"\n",(0,t.jsx)(n.p,{children:"Ci-dessous, nous avons pr\xe9par\xe9 un tutoriel afin de montrer l'utilisation de LLM open-source pour construire un syst\xe8me RAG pour g\xe9n\xe9rer des titres courts et concis d'articles sur l'apprentissage automatique :"}),"\n",(0,t.jsx)(l.oy,{children:(0,t.jsx)(l.Zb,{icon:(0,t.jsx)(d.dN,{}),title:"Commencer avec un RAG",href:"https://github.com/dair-ai/Prompt-Engineering-Guide/blob/main/notebooks/pe-rag.ipynb"})}),"\n",(0,t.jsx)(l.UW,{type:"info",emoji:"\uD83C\uDF93",children:(0,t.jsxs)(n.p,{children:["Vous voulez en savoir plus sur le RAG ? D\xe9couvrez notre ",(0,t.jsx)(n.a,{href:"https://maven.com/dair-ai/prompt-engineering-llms?cohortSlug=",children:"nouveau cours bas\xe9 sur des cohortes"}),". Utilisez le code promo MAVENAI20 pour une r\xe9duction de 20%."]})}),"\n",(0,t.jsx)(n.h2,{id:"r\xe9f\xe9rences",children:"R\xe9f\xe9rences"}),"\n",(0,t.jsx)(n.p,{children:"(en fran\xe7ais en dessous)"}),"\n",(0,t.jsxs)(n.ul,{children:["\n",(0,t.jsxs)(n.li,{children:["\n",(0,t.jsxs)(n.p,{children:[(0,t.jsx)(n.a,{href:"https://arxiv.org/abs/2312.10997",children:"Retrieval-Augmented Generation for Large Language Models: A Survey"})," (Dec 2023)\nFR : ",(0,t.jsx)(n.a,{href:"https://arxiv.org/abs/2312.10997",children:"G\xe9n\xe9ration Augment\xe9e par R\xe9cup\xe9ration pour les Grands Mod\xe8les de Langage : Une synth\xe8se (document EN)"})," (D\xe9c 2023)"]}),"\n"]}),"\n",(0,t.jsxs)(n.li,{children:["\n",(0,t.jsxs)(n.p,{children:[(0,t.jsx)(n.a,{href:"https://ai.meta.com/blog/retrieval-augmented-generation-streamlining-the-creation-of-intelligent-natural-language-processing-models/",children:"Retrieval Augmented Generation: Streamlining the creation of intelligent natural language processing models"})," (Sep 2020)\nFR : ",(0,t.jsx)(n.a,{href:"https://ai.meta.com/blog/retrieval-augmented-generation-streamlining-the-creation-of-intelligent-natural-language-processing-models/",children:"G\xe9n\xe9ration Augment\xe9e par R\xe9cup\xe9ration : Rationaliser la cr\xe9ation de mod\xe8les de traitement naturel du langage intelligents (doc EN)"})," (Sep 2020)"]}),"\n"]}),"\n"]})]})}let p={MDXContent:function(){let e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{},{wrapper:n}=Object.assign({},(0,s.a)(),e.components);return n?(0,t.jsx)(n,{...e,children:(0,t.jsx)(_createMdxContent,{...e})}):_createMdxContent(e)},pageOpts:{filePath:"pages/techniques/rag.fr.mdx",route:"/techniques/rag",timestamp:1726724478e3,pageMap:[{kind:"Meta",locale:"fr",data:{index:"Prompt Engineering",introduction:"Introduction",techniques:"Techniques",applications:"Applications",prompts:"Prompt Hub",models:"Models",risks:"Risques et M\xe9susages",research:"LLM Research Findings",papers:"Papers",tools:"Outils",notebooks:"Notebooks",datasets:"Datasets",readings:"Lectures suppl\xe9mentaires",course:{title:"Prompt Engineering Course",type:"page"},services:{title:"Services",type:"page"},about:{title:"\xc0 propos",type:"page"}}},{kind:"MdxPage",name:"about",route:"/about",locale:"fr"},{kind:"Folder",name:"agents",route:"/agents",children:[{kind:"Meta",locale:"en",data:{introduction:"Introduction to Agents",components:"Agent Components"}},{kind:"MdxPage",name:"components",route:"/agents/components",locale:"en"},{kind:"MdxPage",name:"introduction",route:"/agents/introduction",locale:"en"}]},{kind:"Folder",name:"applications",route:"/applications",children:[{kind:"Meta",locale:"fr",data:{generating:"G\xe9n\xe9ration de donn\xe9es",coding:"Generating Code",workplace_casestudy:"\xc9tude de cas sur la classification des emplois des dipl\xf4m\xe9s",pf:"Prompt Function",function_calling:"Function Calling"}},{kind:"MdxPage",name:"coding",route:"/applications/coding",locale:"fr"},{kind:"MdxPage",name:"function_calling",route:"/applications/function_calling",locale:"fr"},{kind:"MdxPage",name:"generating",route:"/applications/generating",locale:"fr"},{kind:"MdxPage",name:"pf",route:"/applications/pf",locale:"fr"},{kind:"MdxPage",name:"workplace_casestudy",route:"/applications/workplace_casestudy",locale:"fr"},{kind:"MdxPage",name:"context-caching",route:"/applications/context-caching",locale:"en"},{kind:"MdxPage",name:"finetuning-gpt4o",route:"/applications/finetuning-gpt4o",locale:"en"},{kind:"MdxPage",name:"generating_textbooks",route:"/applications/generating_textbooks",locale:"en"},{kind:"MdxPage",name:"synthetic_rag",route:"/applications/synthetic_rag",locale:"en"}]},{kind:"MdxPage",name:"applications",route:"/applications",locale:"fr"},{kind:"MdxPage",name:"course",route:"/course",locale:"fr"},{kind:"MdxPage",name:"datasets",route:"/datasets",locale:"fr"},{kind:"Folder",name:"guides",route:"/guides",children:[{kind:"Meta",locale:"en",data:{"optimizing-prompts":"Optimizing Prompts"}},{kind:"MdxPage",name:"optimizing-prompts",route:"/guides/optimizing-prompts",locale:"en"}]},{kind:"MdxPage",name:"index",route:"/",locale:"fr"},{kind:"Folder",name:"introduction",route:"/introduction",children:[{kind:"Meta",locale:"fr",data:{settings:"Param\xe8tres LLM",basics:"Principes de base en Prompting",elements:"\xc9l\xe9ments d'un prompt",tips:"Conseils g\xe9n\xe9raux pour la conception de prompts",examples:"Exemples de prompts"}},{kind:"MdxPage",name:"basics",route:"/introduction/basics",locale:"fr"},{kind:"MdxPage",name:"elements",route:"/introduction/elements",locale:"fr"},{kind:"MdxPage",name:"examples",route:"/introduction/examples",locale:"fr"},{kind:"MdxPage",name:"settings",route:"/introduction/settings",locale:"fr"},{kind:"MdxPage",name:"tips",route:"/introduction/tips",locale:"fr"}]},{kind:"MdxPage",name:"introduction",route:"/introduction",locale:"fr"},{kind:"Folder",name:"models",route:"/models",children:[{kind:"Meta",locale:"fr",data:{flan:"Flan",chatgpt:"ChatGPT",llama:"LLaMA","gpt-4":"GPT-4","mistral-7b":"Mistral 7B",gemini:"Gemini","gemini-advanced":"Gemini Advanced","gemini-pro":"Gemini 1.5 Pro","phi-2":"Phi-2",mixtral:"Mixtral","code-llama":"Code Llama",olmo:"OLMo",sora:"Sora",collection:"Collection de mod\xe8les"}},{kind:"MdxPage",name:"chatgpt",route:"/models/chatgpt",locale:"fr"},{kind:"MdxPage",name:"code-llama",route:"/models/code-llama",locale:"fr"},{kind:"MdxPage",name:"collection",route:"/models/collection",locale:"fr"},{kind:"MdxPage",name:"flan",route:"/models/flan",locale:"fr"},{kind:"MdxPage",name:"gemini-advanced",route:"/models/gemini-advanced",locale:"fr"},{kind:"MdxPage",name:"gemini-pro",route:"/models/gemini-pro",locale:"fr"},{kind:"MdxPage",name:"gemini",route:"/models/gemini",locale:"fr"},{kind:"MdxPage",name:"gpt-4",route:"/models/gpt-4",locale:"fr"},{kind:"MdxPage",name:"llama",route:"/models/llama",locale:"fr"},{kind:"MdxPage",name:"mistral-7b",route:"/models/mistral-7b",locale:"fr"},{kind:"MdxPage",name:"mixtral",route:"/models/mixtral",locale:"fr"},{kind:"MdxPage",name:"olmo",route:"/models/olmo",locale:"fr"},{kind:"MdxPage",name:"phi-2",route:"/models/phi-2",locale:"fr"},{kind:"MdxPage",name:"sora",route:"/models/sora",locale:"fr"},{kind:"MdxPage",name:"claude-3",route:"/models/claude-3",locale:"en"},{kind:"MdxPage",name:"gemma",route:"/models/gemma",locale:"en"},{kind:"MdxPage",name:"grok-1",route:"/models/grok-1",locale:"en"},{kind:"MdxPage",name:"llama-3",route:"/models/llama-3",locale:"en"},{kind:"MdxPage",name:"mistral-large",route:"/models/mistral-large",locale:"en"},{kind:"MdxPage",name:"mixtral-8x22b",route:"/models/mixtral-8x22b",locale:"en"}]},{kind:"MdxPage",name:"models",route:"/models",locale:"fr"},{kind:"MdxPage",name:"notebooks",route:"/notebooks",locale:"fr"},{kind:"MdxPage",name:"papers",route:"/papers",locale:"fr"},{kind:"Folder",name:"prompts",route:"/prompts",children:[{kind:"Meta",locale:"fr",data:{classification:"Classification",coding:"Coding",creativity:"Creativity",evaluation:"Evaluation","information-extraction":"Information Extraction","image-generation":"Image Generation",mathematics:"Mathematics","question-answering":"Question Answering",reasoning:"Reasoning","text-summarization":"Text Summarization",truthfulness:"Truthfulness","adversarial-prompting":"Adversarial Prompting"}},{kind:"Folder",name:"adversarial-prompting",route:"/prompts/adversarial-prompting",children:[{kind:"Meta",locale:"en",data:{"prompt-injection":"Prompt Injection","prompt-leaking":"Prompt Leaking","jailbreaking-llms":"Jailbreaking"}},{kind:"MdxPage",name:"jailbreaking-llms",route:"/prompts/adversarial-prompting/jailbreaking-llms",locale:"en"},{kind:"MdxPage",name:"prompt-injection",route:"/prompts/adversarial-prompting/prompt-injection",locale:"en"},{kind:"MdxPage",name:"prompt-leaking",route:"/prompts/adversarial-prompting/prompt-leaking",locale:"en"}]},{kind:"Folder",name:"classification",route:"/prompts/classification",children:[{kind:"Meta",locale:"en",data:{sentiment:"Sentiment Classification","sentiment-fewshot":"Few-Shot Sentiment Classification"}},{kind:"MdxPage",name:"sentiment-fewshot",route:"/prompts/classification/sentiment-fewshot",locale:"en"},{kind:"MdxPage",name:"sentiment",route:"/prompts/classification/sentiment",locale:"en"}]},{kind:"Folder",name:"coding",route:"/prompts/coding",children:[{kind:"Meta",locale:"en",data:{"code-snippet":"Generate Code Snippet","mysql-query":"Generate MySQL Query",tikz:"Draw TiKZ Diagram"}},{kind:"MdxPage",name:"code-snippet",route:"/prompts/coding/code-snippet",locale:"en"},{kind:"MdxPage",name:"mysql-query",route:"/prompts/coding/mysql-query",locale:"en"},{kind:"MdxPage",name:"tikz",route:"/prompts/coding/tikz",locale:"en"}]},{kind:"Folder",name:"creativity",route:"/prompts/creativity",children:[{kind:"Meta",locale:"en",data:{rhymes:"Rhymes","infinite-primes":"Infinite Primes",interdisciplinary:"Interdisciplinary","new-words":"Inventing New Words"}},{kind:"MdxPage",name:"infinite-primes",route:"/prompts/creativity/infinite-primes",locale:"en"},{kind:"MdxPage",name:"interdisciplinary",route:"/prompts/creativity/interdisciplinary",locale:"en"},{kind:"MdxPage",name:"new-words",route:"/prompts/creativity/new-words",locale:"en"},{kind:"MdxPage",name:"rhymes",route:"/prompts/creativity/rhymes",locale:"en"}]},{kind:"Folder",name:"evaluation",route:"/prompts/evaluation",children:[{kind:"Meta",locale:"en",data:{"plato-dialogue":"Evaluate Plato's Dialogue"}},{kind:"MdxPage",name:"plato-dialogue",route:"/prompts/evaluation/plato-dialogue",locale:"en"}]},{kind:"Folder",name:"image-generation",route:"/prompts/image-generation",children:[{kind:"Meta",locale:"en",data:{"alphabet-person":"Draw a Person Using Alphabet"}},{kind:"MdxPage",name:"alphabet-person",route:"/prompts/image-generation/alphabet-person",locale:"en"}]},{kind:"Folder",name:"information-extraction",route:"/prompts/information-extraction",children:[{kind:"Meta",locale:"en",data:{"extract-models":"Extract Model Names"}},{kind:"MdxPage",name:"extract-models",route:"/prompts/information-extraction/extract-models",locale:"en"}]},{kind:"Folder",name:"mathematics",route:"/prompts/mathematics",children:[{kind:"Meta",locale:"en",data:{"composite-functions":"Evaluating Composite Functions","odd-numbers":"Adding Odd Numbers"}},{kind:"MdxPage",name:"composite-functions",route:"/prompts/mathematics/composite-functions",locale:"en"},{kind:"MdxPage",name:"odd-numbers",route:"/prompts/mathematics/odd-numbers",locale:"en"}]},{kind:"Folder",name:"question-answering",route:"/prompts/question-answering",children:[{kind:"Meta",locale:"en",data:{"closed-domain":"Closed Domain Question Answering","open-domain":"Open Domain Question Answering","science-qa":"Science Question Answering"}},{kind:"MdxPage",name:"closed-domain",route:"/prompts/question-answering/closed-domain",locale:"en"},{kind:"MdxPage",name:"open-domain",route:"/prompts/question-answering/open-domain",locale:"en"},{kind:"MdxPage",name:"science-qa",route:"/prompts/question-answering/science-qa",locale:"en"}]},{kind:"Folder",name:"reasoning",route:"/prompts/reasoning",children:[{kind:"Meta",locale:"en",data:{"indirect-reasoning":"Indirect Reasoning","physical-reasoning":"Physical Reasoning"}},{kind:"MdxPage",name:"indirect-reasoning",route:"/prompts/reasoning/indirect-reasoning",locale:"en"},{kind:"MdxPage",name:"physical-reasoning",route:"/prompts/reasoning/physical-reasoning",locale:"en"}]},{kind:"Folder",name:"text-summarization",route:"/prompts/text-summarization",children:[{kind:"Meta",locale:"en",data:{"explain-concept":"Explain A Concept"}},{kind:"MdxPage",name:"explain-concept",route:"/prompts/text-summarization/explain-concept",locale:"en"}]},{kind:"Folder",name:"truthfulness",route:"/prompts/truthfulness",children:[{kind:"Meta",locale:"en",data:{"identify-hallucination":"Hallucination Identification"}},{kind:"MdxPage",name:"identify-hallucination",route:"/prompts/truthfulness/identify-hallucination",locale:"en"}]},{kind:"MdxPage",name:"adversarial-prompting",route:"/prompts/adversarial-prompting",locale:"en"},{kind:"MdxPage",name:"classification",route:"/prompts/classification",locale:"en"},{kind:"MdxPage",name:"coding",route:"/prompts/coding",locale:"en"},{kind:"MdxPage",name:"creativity",route:"/prompts/creativity",locale:"en"},{kind:"MdxPage",name:"evaluation",route:"/prompts/evaluation",locale:"en"},{kind:"MdxPage",name:"image-generation",route:"/prompts/image-generation",locale:"en"},{kind:"MdxPage",name:"information-extraction",route:"/prompts/information-extraction",locale:"en"},{kind:"MdxPage",name:"mathematics",route:"/prompts/mathematics",locale:"en"},{kind:"MdxPage",name:"question-answering",route:"/prompts/question-answering",locale:"en"},{kind:"MdxPage",name:"reasoning",route:"/prompts/reasoning",locale:"en"},{kind:"MdxPage",name:"text-summarization",route:"/prompts/text-summarization",locale:"en"},{kind:"MdxPage",name:"truthfulness",route:"/prompts/truthfulness",locale:"en"}]},{kind:"MdxPage",name:"prompts",route:"/prompts",locale:"fr"},{kind:"MdxPage",name:"readings",route:"/readings",locale:"fr"},{kind:"Folder",name:"research",route:"/research",children:[{kind:"Meta",locale:"fr",data:{"llm-agents":"LLM Agents",rag:"RAG for LLMs","trustworthiness-in-llms":"Trustworthiness in LLMs"}},{kind:"MdxPage",name:"llm-agents",route:"/research/llm-agents",locale:"fr"},{kind:"MdxPage",name:"rag",route:"/research/rag",locale:"fr"},{kind:"MdxPage",name:"trustworthiness-in-llms",route:"/research/trustworthiness-in-llms",locale:"fr"},{kind:"MdxPage",name:"groq",route:"/research/groq",locale:"en"},{kind:"MdxPage",name:"guided-cot",route:"/research/guided-cot",locale:"en"},{kind:"MdxPage",name:"infini-attention",route:"/research/infini-attention",locale:"en"},{kind:"MdxPage",name:"llm-reasoning",route:"/research/llm-reasoning",locale:"en"},{kind:"MdxPage",name:"llm-recall",route:"/research/llm-recall",locale:"en"},{kind:"MdxPage",name:"llm-tokenization",route:"/research/llm-tokenization",locale:"en"},{kind:"MdxPage",name:"rag-faithfulness",route:"/research/rag-faithfulness",locale:"en"},{kind:"MdxPage",name:"rag_hallucinations",route:"/research/rag_hallucinations",locale:"en"},{kind:"MdxPage",name:"synthetic_data",route:"/research/synthetic_data",locale:"en"},{kind:"MdxPage",name:"thoughtsculpt",route:"/research/thoughtsculpt",locale:"en"}]},{kind:"MdxPage",name:"research",route:"/research",locale:"fr"},{kind:"Folder",name:"risks",route:"/risks",children:[{kind:"Meta",locale:"fr",data:{adversarial:"Incitation contradictoire",factuality:"Factualit\xe9",biases:"Biases"}},{kind:"MdxPage",name:"adversarial",route:"/risks/adversarial",locale:"fr"},{kind:"MdxPage",name:"biases",route:"/risks/biases",locale:"fr"},{kind:"MdxPage",name:"factuality",route:"/risks/factuality",locale:"fr"}]},{kind:"MdxPage",name:"risks",route:"/risks",locale:"fr"},{kind:"MdxPage",name:"services",route:"/services",locale:"fr"},{kind:"Folder",name:"techniques",route:"/techniques",children:[{kind:"Meta",locale:"fr",data:{zeroshot:"Zero-shot Prompting",fewshot:"Few-shot Prompting",cot:"Chain-of-Thought Prompting",consistency:"Self-Consistency",knowledge:"Generate Knowledge Prompting",prompt_chaining:"Prompt Chaining",tot:"Tree of Thoughts",rag:"Retrieval Augmented Generation",art:"Automatic Reasoning and Tool-use",ape:"Automatic Prompt Engineer",activeprompt:"Active-Prompt",dsp:"Directional Stimulus Prompting",pal:"Program-Aided Language Models",react:"ReAct",reflexion:"Reflexion",multimodalcot:"Multimodal CoT",graph:"Graph Prompting"}},{kind:"MdxPage",name:"activeprompt",route:"/techniques/activeprompt",locale:"fr"},{kind:"MdxPage",name:"ape",route:"/techniques/ape",locale:"fr"},{kind:"MdxPage",name:"art",route:"/techniques/art",locale:"fr"},{kind:"MdxPage",name:"consistency",route:"/techniques/consistency",locale:"fr"},{kind:"MdxPage",name:"cot",route:"/techniques/cot",locale:"fr"},{kind:"MdxPage",name:"dsp",route:"/techniques/dsp",locale:"fr"},{kind:"MdxPage",name:"fewshot",route:"/techniques/fewshot",locale:"fr"},{kind:"MdxPage",name:"graph",route:"/techniques/graph",locale:"fr"},{kind:"MdxPage",name:"knowledge",route:"/techniques/knowledge",locale:"fr"},{kind:"MdxPage",name:"multimodalcot",route:"/techniques/multimodalcot",locale:"fr"},{kind:"MdxPage",name:"pal",route:"/techniques/pal",locale:"fr"},{kind:"MdxPage",name:"prompt_chaining",route:"/techniques/prompt_chaining",locale:"fr"},{kind:"MdxPage",name:"rag",route:"/techniques/rag",locale:"fr"},{kind:"MdxPage",name:"react",route:"/techniques/react",locale:"fr"},{kind:"MdxPage",name:"reflexion",route:"/techniques/reflexion",locale:"fr"},{kind:"MdxPage",name:"tot",route:"/techniques/tot",locale:"fr"},{kind:"MdxPage",name:"zeroshot",route:"/techniques/zeroshot",locale:"fr"},{kind:"MdxPage",name:"meta-prompting",route:"/techniques/meta-prompting",locale:"en"}]},{kind:"MdxPage",name:"techniques",route:"/techniques",locale:"fr"},{kind:"MdxPage",name:"tools",route:"/tools",locale:"fr"},{kind:"MdxPage",name:"agents",route:"/agents",locale:"en"}],flexsearch:{codeblocks:!0},title:"G\xe9n\xe9ration Augment\xe9e par R\xe9cup\xe9ration (RAG)",headings:u},pageNextRoute:"/techniques/rag.fr",nextraLayout:o.ZP,themeConfig:r.Z};n.default=(0,i.j)(p)},94699:function(e,n,a){"use strict";a.d(n,{w:function(){return Screenshot}});var t=a(11527),i=a(86259),o=a(87858),r=a.n(o);function Screenshot(e){let{src:n,alt:a,full:o}=e;return(0,t.jsx)("div",{className:(0,i.Z)("mt-6 -mb-4 flex justify-center overflow-hidden rounded-xl border dark:border-zinc-800",o?"bg-white":"bg-zinc-100"),children:(0,t.jsx)(r(),{src:n,alt:a,className:(0,i.Z)("w-auto select-none bg-white",o?"":"ring-1 ring-gray-200")})})}}},function(e){e.O(0,[63028,72767,49774,92888,40179],function(){return e(e.s=84798)}),_N_E=e.O()}]);