{"id":226450,"date":"2024-02-13T06:30:00","date_gmt":"2024-02-13T06:30:00","guid":{"rendered":"https:\/\/datascientest.com\/?p=226450"},"modified":"2024-03-14T15:05:09","modified_gmt":"2024-03-14T15:05:09","slug":"advanced-prompt-engineering-tout-savoir","status":"publish","type":"post","link":"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir","title":{"rendered":"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"226450\" class=\"elementor elementor-226450\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-e838e17 e-flex e-con-boxed e-con\" data-id=\"e838e17\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-6cfbfd4 elementor-widget elementor-widget-heading\" data-id=\"6cfbfd4\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><h2 class=\"elementor-heading-title elementor-size-default\">Avec la d\u00e9mocratisation de l\u2019intelligence artificielle et des large language models LLM ( ou grands mod\u00e8les de langage), la r\u00e9solution de probl\u00e8me a \u00e9t\u00e9 boulevers\u00e9e. Pour de nombreuses probl\u00e9matiques, il n\u2019est plus n\u00e9cessaire de cr\u00e9er un programme informatique avec un langage de programmation complexe. \u00c0 la place, il suffit d\u2019un bon prompt textuel. <\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-9b8e8ed elementor-widget elementor-widget-text-editor\" data-id=\"9b8e8ed\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Et plus pr\u00e9cis\u00e9ment de l\u2019advanced prompt engineering qui red\u00e9finit la mani\u00e8re dont nous optimisons les performances des mod\u00e8les de langage, tels que le <\/span><a href=\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\"><span style=\"font-weight: 400;\">ChatGPT.\u00a0<\/span><\/a><\/p><p><span style=\"font-weight: 400;\">D\u00e9couvrez l\u2019int\u00e9r\u00eat de l\u2019ing\u00e9nierie des prompts avanc\u00e9s \u00e0 l\u2019\u00e8re de l\u2019IA, avec en prime 4 exemples concrets de techniques \u00e0 conna\u00eetre.\u00a0<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-0096c5e e-flex e-con-boxed e-con\" data-id=\"0096c5e\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-d4f8636 elementor-widget elementor-widget-heading\" data-id=\"d4f8636\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">L\u2019Advanced Prompt Engineering, une approche sophistiqu\u00e9e de l\u2019utilisation des LLM<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-529d556 elementor-widget elementor-widget-text-editor\" data-id=\"529d556\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">L&rsquo;advanced prompt engineering fait r\u00e9f\u00e9rence au <\/span><b>processus d&rsquo;affinage de l&rsquo;entr\u00e9e d&rsquo;un mod\u00e8le de langage afin d&rsquo;obtenir la sortie souhait\u00e9e<\/b><span style=\"font-weight: 400;\"> sans modifier les poids r\u00e9els du mod\u00e8le. Il s\u2019agit d\u2019une approche sophistiqu\u00e9e qui va au-del\u00e0 des ajustements classiques des poids du mod\u00e8le. Ici, on se concentre sur le raffinement habile de l&rsquo;entr\u00e9e pour influencer la sortie d\u00e9sir\u00e9e.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Par exemple, pour r\u00e9soudre certains probl\u00e8mes sur <a href=\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\">GPT-3<\/a>, une simple phrase a \u00e9t\u00e9 ajout\u00e9e :<\/span><b> \u00ab\u00a0R\u00e9fl\u00e9chissons \u00e9tape par \u00e9tape\u00a0\u00bb. <\/b><span style=\"font-weight: 400;\">Les r\u00e9sultats sont bluffants ! Le taux de r\u00e9solution de GPT-3 est pass\u00e9 de 18 % \u00e0 79 %.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">En <\/span><b>\u00e9laborant des prompts bien structur\u00e9s,<\/b><span style=\"font-weight: 400;\"> vous pouvez guider les mod\u00e8les de langage pour produire des r\u00e9sultats pr\u00e9cis et coh\u00e9rents. Mais attention, car pour exploiter pleinement leur potentiel, il ne suffit pas de poser de simples questions. Il est n\u00e9cessaire d\u2019optimiser les entr\u00e9es, avec des instructions claires, pour <\/span><b>influencer le comportement du mod\u00e8le et garantir des r\u00e9sultats pertinents.<\/b><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-fd21c74 elementor-widget elementor-widget-image\" data-id=\"fd21c74\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1000\" height=\"667\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering1.jpg\" class=\"attachment-full size-full wp-image-226460\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering1.jpg 1000w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering1-300x200.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering1-768x512.jpg 768w\" sizes=\"(max-width: 1000px) 100vw, 1000px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-4fa5cea elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"4fa5cea\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-prompt-engineering\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Apprendre \u00e0 utiliser l'Advanced Prompt Engineering<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-55bf771 e-flex e-con-boxed e-con\" data-id=\"55bf771\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-eb29ea7 elementor-widget elementor-widget-heading\" data-id=\"eb29ea7\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">4 techniques avanc\u00e9es d\u2019ing\u00e9nierie de prompt<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7f02ca5 elementor-widget elementor-widget-text-editor\" data-id=\"7f02ca5\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Ma\u00eetriser l\u2019advanced prompt engineering signifie conna\u00eetre les diff\u00e9rentes techniques de messages guid\u00e9s pour obtenir le r\u00e9sultat souhait\u00e9. Voici les techniques les plus courantes. <\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-89743de elementor-widget elementor-widget-heading\" data-id=\"89743de\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-medium\">Zero shots Prompt<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-923244f elementor-widget elementor-widget-text-editor\" data-id=\"923244f\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Le Zero Shots Prompt est l\u2019une des premi\u00e8res techniques de l\u2019advanced Prompt engineering. Il s\u2019agit simplement de <\/span><b>poser une question ou de pr\u00e9senter une t\u00e2che sp\u00e9cifique <\/b><span style=\"font-weight: 400;\">au mod\u00e8le de langage. Et ce,<\/span><b> sans que le mod\u00e8le n\u2019ait \u00e9t\u00e9 sp\u00e9cifiquement pr\u00e9-entra\u00een\u00e9<\/b><span style=\"font-weight: 400;\"> \u00e0 la t\u00e2che ou la question. Autrement dit, il n\u2019a jamais \u00e9t\u00e9 expos\u00e9 \u00e0 des exemples sp\u00e9cifiques relatifs \u00e0 cette t\u00e2che au cours de sa phase d&rsquo;entra\u00eenement.<\/span><\/p><p><span style=\"font-weight: 400;\">Cette approche est souvent utilis\u00e9e pour<\/span><b> \u00e9valuer la capacit\u00e9 de g\u00e9n\u00e9ralisation d&rsquo;un mod\u00e8le de langage. <\/b><span style=\"font-weight: 400;\">Et pour cause, avec le Zero shots prompt, il d\u00e9montre qu\u2019il est capable de r\u00e9pondre \u00e0 des requ\u00eates pour lesquelles il n&rsquo;a pas \u00e9t\u00e9 explicitement programm\u00e9 pendant son apprentissage. <\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-af38ade elementor-widget elementor-widget-heading\" data-id=\"af38ade\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-medium\">Chain of thought prompting (CoT)<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-5373706 elementor-widget elementor-widget-text-editor\" data-id=\"5373706\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Chain of Thought sugg\u00e8re une s\u00e9quence de pens\u00e9es ou d&rsquo;id\u00e9es interconnect\u00e9es. Appliqu\u00e9 au machine learning, le CoT permet d\u2019am\u00e9liorer les performances du mod\u00e8le sur des t\u00e2ches de raisonnement plus complexes.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Concr\u00e8tement, les ing\u00e9nieurs \u00e0 l\u2019origine de cette technique d\u2019advanced prompt engineering ont \u00e9tudi\u00e9 la g\u00e9n\u00e9ration d&rsquo;une cha\u00eene de pens\u00e9e humaine. Et cela passe par une s\u00e9rie d&rsquo;\u00e9tapes de raisonnement interm\u00e9diaires.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Pour appliquer ces principes aux mod\u00e8les de langage, il convient de leur fournir quelques d\u00e9monstrations de cha\u00eenes de pens\u00e9e en tant qu&rsquo;exemples. Ensuite, il faut leur demander de <\/span><b>produire quelques \u00e9tapes de raisonnement <\/b><span style=\"font-weight: 400;\">avant de fournir la r\u00e9ponse finale. Ce qui leur permet d\u2019effectuer des raisonnements complexes. <\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-68b471a elementor-widget elementor-widget-image\" data-id=\"68b471a\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1000\" height=\"667\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering2.jpg\" class=\"attachment-full size-full wp-image-226461\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering2.jpg 1000w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering2-300x200.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering2-768x512.jpg 768w\" sizes=\"(max-width: 1000px) 100vw, 1000px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8364fe6 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"8364fe6\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-prompt-engineering\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Se former aux techniques d'Advanced Prompt Engineering<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-05a763f elementor-widget elementor-widget-heading\" data-id=\"05a763f\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-medium\">La self consistency<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6847bbc elementor-widget elementor-widget-text-editor\" data-id=\"6847bbc\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">En science et en ing\u00e9nierie, la self consistency signifie que les diff\u00e9rentes parties ou composants d\u2019un syst\u00e8me interagissent de mani\u00e8re \u00e0 maintenir leur coh\u00e9rence interne.<\/span><\/p><p><span style=\"font-weight: 400;\">En <a href=\"https:\/\/datascientest.com\/machine-learning-tout-savoir\">machine learning<\/a>, la self consistency est une approche am\u00e9lior\u00e9e de la CoT pour r\u00e9soudre des probl\u00e8mes de raisonnement encore plus complexes.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">\u00c0 travers cette technique d\u2019advanced prompt engineering, les chemins de raisonnement sont \u00e9chantillonn\u00e9s. Ce qui permet au mod\u00e8le de langage de <\/span><b>s\u00e9lectionner la r\u00e9ponse la plus coh\u00e9rente<\/b><span style=\"font-weight: 400;\"> (sans prendre uniquement en compte le chemin le plus rapide).<\/span><\/p><p><span style=\"font-weight: 400;\">Cette technique de prompting est particuli\u00e8rement adapt\u00e9e aux probl\u00e8mes de raisonnement complexe o\u00f9 plusieurs modes de pens\u00e9e existent pour conduire \u00e0 une r\u00e9ponse correcte.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7c6b87d elementor-widget elementor-widget-heading\" data-id=\"7c6b87d\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-medium\">ReAct<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2e2dc8b elementor-widget elementor-widget-text-editor\" data-id=\"2e2dc8b\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">ReAct est une technique d\u2019advanced prompt engineering d\u00e9velopp\u00e9e en 2022. Il s\u2019agit de la <\/span><b>synergie entre Reasoning (raisonnement) et Acting (action).\u00a0<\/b><\/p><p><span style=\"font-weight: 400;\">Cette m\u00e9thode implique que les <a href=\"https:\/\/datascientest.com\/large-language-models-tout-savoir\">LLM<\/a> g\u00e9n\u00e8rent \u00e0 la fois des raisonnements et des actions sp\u00e9cifiques \u00e0 une t\u00e2che. Les raisonnements aident ainsi le mod\u00e8le \u00e0 g\u00e9n\u00e9rer un plan d&rsquo;action. De leur c\u00f4t\u00e9, les actions facilitent l\u2019acc\u00e8s \u00e0 des sources externes de connaissances. Ce qui permet de collecter des informations suppl\u00e9mentaires, et donc d\u2019am\u00e9liorer le raisonnement des LLM.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Avec ReAct, raisonnement et action sont intimement entrelac\u00e9s pour am\u00e9liorer toujours plus les r\u00e9sultats du mod\u00e8le de langage.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-e9633ad e-flex e-con-boxed e-con\" data-id=\"e9633ad\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-e90e153 elementor-widget elementor-widget-heading\" data-id=\"e90e153\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">La ma\u00eetrise de l\u2019advanced prompt engineering avec DataScientest<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-bbceee1 elementor-widget elementor-widget-text-editor\" data-id=\"bbceee1\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<span style=\"font-weight: 400;\">\u00c0 l\u2019heure o\u00f9 les <a href=\"https:\/\/datascientest.com\/intelligence-artificielle-definition\">solutions bas\u00e9es sur l\u2019IA g\u00e9n\u00e9rative<\/a> se multiplient, la ma\u00eetrise de l\u2019advanced prompt engineering se r\u00e9v\u00e8le indispensable. Notamment pour les experts de la data science. Pour am\u00e9liorer vos mod\u00e8les de langage, vous devrez utiliser l\u2019ing\u00e9nierie avanc\u00e9e des messages guid\u00e9s. Mais pas seulement. D\u2019autres techniques sont \u00e0 conna\u00eetre pour cr\u00e9er des LLM performants. C\u2019est pour cette raison qu\u2019il convient de <a href=\"https:\/\/datascientest.com\/formation-data-scientist\">se former avec DataScientest<\/a>.<\/span>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2c80527 elementor-widget elementor-widget-image\" data-id=\"2c80527\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1000\" height=\"667\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering3.jpg\" class=\"attachment-full size-full wp-image-226466\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering3.jpg 1000w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering3-300x200.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/advanced_prompt_engineering3-768x512.jpg 768w\" sizes=\"(max-width: 1000px) 100vw, 1000px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-553c08f elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"553c08f\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-prompt-engineering\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Se former \u00e0 l'Advanced Prompt Engineering<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-ee4e595 elementor-widget elementor-widget-html\" data-id=\"ee4e595\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<script type=\"application\/ld+json\">\r\n{\r\n  \"@context\": \"https:\/\/schema.org\",\r\n  \"@type\": \"FAQPage\",\r\n  \"mainEntity\": [{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Qu\u2019est-ce que l'Advanced Prompt Engineering ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"L'advanced prompt engineering fait r\u00e9f\u00e9rence au processus d'affinage de l'entr\u00e9e d'un mod\u00e8le de langage afin d'obtenir la sortie souhait\u00e9e sans modifier les poids r\u00e9els du mod\u00e8le. Il s\u2019agit d\u2019une approche sophistiqu\u00e9e qui va au-del\u00e0 des ajustements classiques des poids du mod\u00e8le. Ici, on se concentre sur le raffinement habile de l'entr\u00e9e pour influencer la sortie d\u00e9sir\u00e9e. Par exemple, pour r\u00e9soudre certains probl\u00e8mes sur GPT-3, une simple phrase a \u00e9t\u00e9 ajout\u00e9e : \"R\u00e9fl\u00e9chissons \u00e9tape par \u00e9tape\". Les r\u00e9sultats sont bluffants ! Le taux de r\u00e9solution de GPT-3 est pass\u00e9 de 18 % \u00e0 79 %.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Quelles sont les 4 techniques avanc\u00e9es d\u2019ing\u00e9nierie de prompt ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Ma\u00eetriser l\u2019advanced prompt engineering signifie conna\u00eetre les diff\u00e9rentes techniques de messages guid\u00e9s pour obtenir le r\u00e9sultat souhait\u00e9. Voici les techniques les plus courantes : Zero shots Prompt, Chain of thought prompting (CoT), self consistency, ReAct.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"L\u2019advanced prompt engineering et la Data Science\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"\u00c0 l\u2019heure o\u00f9 les solutions bas\u00e9es sur l\u2019IA g\u00e9n\u00e9rative se multiplient, la ma\u00eetrise de l\u2019advanced prompt engineering se r\u00e9v\u00e8le indispensable. Notamment pour les experts de la data science. Pour am\u00e9liorer vos mod\u00e8les de langage, vous devrez utiliser l\u2019ing\u00e9nierie avanc\u00e9e des messages guid\u00e9s. Mais pas seulement. D\u2019autres techniques sont \u00e0 conna\u00eetre pour cr\u00e9er des LLM performants.\"\r\n    }\r\n  }]\r\n}\r\n<\/script>\r\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Avec la d\u00e9mocratisation de l\u2019intelligence artificielle et des large language models LLM ( ou grands mod\u00e8les de langage), la r\u00e9solution de probl\u00e8me a \u00e9t\u00e9 boulevers\u00e9e. Pour de nombreuses probl\u00e9matiques, il n\u2019est plus n\u00e9cessaire de cr\u00e9er un programme informatique avec un langage de programmation complexe. \u00c0 la place, il suffit d\u2019un bon prompt textuel. Et plus [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":226452,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2500],"tags":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?<\/title>\n<meta name=\"description\" content=\"D\u00e9couvrez comment obtenir les r\u00e9sultats souhait\u00e9s sur les outils d\u2019IA g\u00e9n\u00e9rative avec les techniques d\u2019advanced prompt engineering.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2024-02-13T06:30:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-03-14T15:05:09+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/Advanced_Prompt_Engineering.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1050\" \/>\n\t<meta property=\"og:image:height\" content=\"600\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"J\u00e9r\u00e9my Robert\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"J\u00e9r\u00e9my Robert\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir\",\"url\":\"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir\",\"name\":\"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2024-02-13T06:30:00+00:00\",\"dateModified\":\"2024-03-14T15:05:09+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"D\u00e9couvrez comment obtenir les r\u00e9sultats souhait\u00e9s sur les outils d\u2019IA g\u00e9n\u00e9rative avec les techniques d\u2019advanced prompt engineering.\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\",\"name\":\"J\u00e9r\u00e9my Robert\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"caption\":\"J\u00e9r\u00e9my Robert\"},\"description\":\"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/jeremyrobert1\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/jeremyrobert\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?","description":"D\u00e9couvrez comment obtenir les r\u00e9sultats souhait\u00e9s sur les outils d\u2019IA g\u00e9n\u00e9rative avec les techniques d\u2019advanced prompt engineering.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir","og_locale":"fr_FR","og_type":"article","og_title":"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?","og_url":"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2024-02-13T06:30:00+00:00","article_modified_time":"2024-03-14T15:05:09+00:00","og_image":[{"width":1050,"height":600,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/12\/Advanced_Prompt_Engineering.jpg","type":"image\/jpeg"}],"author":"J\u00e9r\u00e9my Robert","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"J\u00e9r\u00e9my Robert","Dur\u00e9e de lecture estim\u00e9e":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir","url":"https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir","name":"Advanced Prompt Engineering : En quoi \u00e7\u00e0 consiste ?","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2024-02-13T06:30:00+00:00","dateModified":"2024-03-14T15:05:09+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"D\u00e9couvrez comment obtenir les r\u00e9sultats souhait\u00e9s sur les outils d\u2019IA g\u00e9n\u00e9rative avec les techniques d\u2019advanced prompt engineering.","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/advanced-prompt-engineering-tout-savoir"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925","name":"J\u00e9r\u00e9my Robert","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","caption":"J\u00e9r\u00e9my Robert"},"description":"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.","sameAs":["https:\/\/www.linkedin.com\/in\/jeremyrobert1\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/jeremyrobert"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/226450"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=226450"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/226450\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/226452"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=226450"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=226450"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=226450"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}