fixed dynamic cards

pull/357/head
Elvis Saravia 4 months ago
parent 4271d1fc3c
commit 5fa029115d

@ -0,0 +1,29 @@
// components/ContentFileNames.tsx
import React, { useEffect, useState } from 'react';
import { Cards, Card } from 'nextra-theme-docs';
import { FilesIcon } from './icons';
const ContentFileNames = ({ section = 'research', lang = 'en' }) => {
const [fileNames, setFileNames] = useState([]);
useEffect(() => {
fetch(`/api/contentFiles?section=${section}&lang=${lang}`)
.then(response => response.json())
.then(data => setFileNames(data.fileNames));
}, [section, lang]);
return (
<Cards>
{fileNames.map(({ slug, title }, index) => (
<Card
key={index}
icon={<FilesIcon />}
title={title}
href={`/${section}/${slug}`}
/>
))}
</Cards>
);
};
export default ContentFileNames;

@ -0,0 +1,27 @@
// pages/api/contentFiles.js
import fs from 'fs';
import path from 'path';
export default function handler(req, res) {
const { section = 'research', lang = 'en' } = req.query;
const directoryPath = path.join(process.cwd(), 'pages', section);
const metaFilePath = path.join(directoryPath, `_meta.${lang}.json`);
let titles = {};
let fileNames = [];
if (fs.existsSync(metaFilePath)) {
const metaFileContents = fs.readFileSync(metaFilePath, 'utf8');
titles = JSON.parse(metaFileContents);
// Iterate over the keys in the titles object to maintain order
fileNames = Object.keys(titles).map(slug => {
return {
slug,
title: titles[slug]
};
});
}
res.status(200).json({ fileNames });
}

@ -1,9 +1,8 @@
# Aplicacions de Prompts
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
En aquesta secció, tractarem algunes maneres avançades i interessants d'utilitzar l'enginyeria de prompts per realitzar tasques útils i més avançades.
<Callout emoji="⚠️">
Aquesta secció està en plena fase de desenvolupament.
</Callout>
<ContentFileNames section="applications" lang="ca"/>

@ -1,46 +1,10 @@
# LLM-Anwendungen
import { Callout } from 'nextra-theme-docs';
import { Cards, Card } from 'nextra-theme-docs';
import { FilesIcon } from 'components/icons';
import ContentFileNames from 'components/ContentFileNames'
In diesem Abschnitt werden wir einige fortgeschrittene und interessante Methoden besprechen, wie wir Prompt-Engineering nutzen können, um nützliche und anspruchsvollere Aufgaben mit LLMs (große Sprachmodelle) zu bewältigen.
<Cards>
<Card
icon={<FilesIcon />}
title="Funktionsaufrufe"
href="/applications/function_calling"
/>
<Card
icon={<FilesIcon />}
title="Generierung von Daten"
href="/applications/generating"
/>
<Card
icon={<FilesIcon />}
title="Generierung eines synthetischen Datensatzes für RAG"
href="/applications/synthetic_rag"
/>
<Card
icon={<FilesIcon />}
title="Umgang mit generierten Datensätzen und deren Vielfalt"
href="/applications/generating_textbooks"
/>
<Card
icon={<FilesIcon />}
title="Codegenerierung"
href="/applications/coding"
/>
<Card
icon={<FilesIcon />}
title="Fallstudie zur Klassifizierung von Absolventenjobs"
href="/applications/workplace_casestudy"
/>
<Card
icon={<FilesIcon />}
title="Prompt-Funktion"
href="/applications/pf"
/>
</Cards>
<ContentFileNames section="applications" lang="de"/>

@ -1,46 +1,10 @@
# LLM Applications
import { Callout } from 'nextra-theme-docs'
import {Cards, Card} from 'nextra-theme-docs'
import {FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
In this section, we will cover advanced and interesting ways we can use prompt engineering to perform useful and more advanced tasks with large language models (LLMs).
<Cards>
<Card
icon={<FilesIcon />}
title="Function Calling"
href="/applications/function_calling"
/>
<Card
icon={<FilesIcon />}
title="Generating Data"
href="/applications/generating"
/>
<Card
icon={<FilesIcon />}
title="Generating Synthetic Dataset for RAG"
href="/applications/synthetic_rag"
/>
<Card
icon={<FilesIcon />}
title="Tackling Generated Datasets Diversity"
href="/applications/generating_textbooks"
/>
<Card
icon={<FilesIcon />}
title="Generating Code"
href="/applications/coding"
/>
<Card
icon={<FilesIcon />}
title="Graduate Job Classification"
href="/applications/workplace_casestudy"
/>
<Card
icon={<FilesIcon />}
title="Prompt Function"
href="/applications/pf"
/>
</Cards>
<ContentFileNames section="applications" lang="en"/>

@ -1,9 +1,8 @@
# Aplicaciones del Prompting
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
En esta sección se mostrarán algunas formas avanzadas e interesantes en las que podemos usar la ingenieria de prompts para realizar tareas más avanzadas y útiles.
<Callout emoji="⚠️">
Esta sección está en pleno desarrollo.
</Callout>
<ContentFileNames section="applications" lang="es"/>

@ -1,9 +1,8 @@
# Kehottesovellukset
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Tässä osiossa käsitellään joitakin edistyneitä ja mielenkiintoisia menetelmiä, joiden avulla voimme soveltaa kehotteita käytännöllisiin ja vaativiin tehtäviin.
<Callout emoji="⚠️">
Tämä osa sivustoa kehittyy jatkuvasti.
</Callout>
<ContentFileNames section="applications" lang="fi"/>

@ -1,9 +1,8 @@
# Prompting Applications
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Dans cette section, nous aborderons certaines façons avancées et intéressantes d'utiliser le prompt engineering pour effectuer des tâches utiles et plus avancées.
<Callout emoji="⚠️">
Cette section est en plein développement.
</Callout>
<ContentFileNames section="applications" lang="fr"/>

@ -1,9 +1,8 @@
# Applicazioni di Prompting
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
In questa sezione tratteremo alcuni modi avanzati e interessanti per utilizzare il prompt engineering per eseguire compiti utili e più avanzati.
<Callout emoji="⚠️">
Questa sezione è in forte sviluppo.
</Callout>
<ContentFileNames section="applications" lang="it"/>

@ -1,9 +1,8 @@
# プロンプトアプリケーション
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
このガイドでは、プロンプトエンジニアリングを使って便利でより高度なタスクを実行するための、高度で興味深い方法について説明します。
<Callout emoji="⚠️">
このセクションは、現在開発が進んでいます。
</Callout>
<ContentFileNames section="applications" lang="jp"/>

@ -1,9 +1,9 @@
# Prompting Applications
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
이 장에서는 프롬프트 엔지니어링을 사용하여 유용한 고급 작업을 수행할 수 있는 몇 가지 흥미로운 고급 방법을 다룹니다.
<Callout emoji="⚠️">
해당 페이지는 개발 중에 있습니다.
</Callout>
<ContentFileNames section="applications" lang="kr"/>

@ -1,9 +1,8 @@
# Prompting e Aplicativos
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Nesta seção, abordaremos algumas maneiras avançadas e interessantes de usar a engenharia de prompt para executar tarefas úteis e mais avançadas.
<Callout emoji="⚠️">
Esta seção está em intenso desenvolvimento.
</Callout>
<ContentFileNames section="applications" lang="pt"/>

@ -1,9 +1,8 @@
# Применение промптов
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
В этом разделе мы рассмотрим некоторые продвинутые и интересные способы использования инженерии промптов для выполнения полезных и более сложных задач.
<Callout emoji="⚠️">
Этот раздел находится в активной разработке.
</Callout>
<ContentFileNames section="applications" lang="ru"/>

@ -1,9 +1,8 @@
# İstemci Uygulamaları
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Bu bölümde, yararlı ve daha gelişmiş görevleri gerçekleştirmek için hızlı mühendisliği kullanabileceğimiz bazı gelişmiş ve ilginç yolları ele alacağız.
<Callout emoji="⚠️">
Bu bölüm yoğun bir geliştirme aşamasındadır.
</Callout>
<ContentFileNames section="applications" lang="tr"/>

@ -1,9 +1,9 @@
# 提示应用
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
在本指南中,我们将介绍一些高级和有趣的方法,利用提示工程来执行有用和更高级的任务。
<Callout emoji="⚠️">
本节正在大力开发中。
</Callout>
<ContentFileNames section="applications" lang="zh"/>

@ -1,8 +1,11 @@
# Introduction
import ContentFileNames from 'components/ContentFileNames'
Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large language models (LLMs). Researchers use prompt engineering to improve the capacity of LLMs on a wide range of common and complex tasks such as question answering and arithmetic reasoning. Developers use prompt engineering to design robust and effective prompting techniques that interface with LLMs and other tools.
This guide covers the basics of prompts to provide a rough idea of how to use prompts to interact and instruct LLMs.
All examples are tested with `text-davinci-003` using [OpenAI's playground](https://platform.openai.com/playground) unless otherwise specified. The model uses the default configurations, i.e., `temperature=0.7` and `top-p=1`.
<ContentFileNames section="introduction" lang="ca"/>

@ -7,6 +7,8 @@ import {
WarningIcon,
FilesIcon,
} from 'components/icons';
import ContentFileNames from 'components/ContentFileNames'
Prompt-Engineering ist eine relativ neue Disziplin, die sich mit der Entwicklung und Optimierung von Prompts beschäftigt, um große Sprachmodelle (LLMs) effizient für eine Vielzahl von Anwendungen und Einsatzmöglichkeiten zu nutzen und zu entwickeln.
@ -16,30 +18,4 @@ Dieser umfassende Leitfaden behandelt die Theorie und praktischen Aspekte des Pr
Alle Beispiele wurden mit `gpt-3.5-turbo` unter Verwendung von [OpenAIs Playground](https://platform.openai.com/playground) getestet, sofern nicht anders angegeben. Das Modell verwendet die Standardeinstellungen, d.h., `temperature=0.7` und `top-p=1`. Die Prompts sollten auch mit anderen Modellen funktionieren, die ähnliche Fähigkeiten wie `gpt-3.5-turbo` haben, aber es könnten sich vollkommen andere Ergebnisse ergeben.
<Cards>
<Card
icon={<FilesIcon />}
title="LLM Einstellungen"
href="/introduction/settings"
/>
<Card
icon={<FilesIcon />}
title="Grundlagen des Promptings"
href="/introduction/basics"
/>
<Card
icon={<FilesIcon />}
title="Elemente eines Prompts"
href="/introduction/elements"
/>
<Card
icon={<FilesIcon />}
title="Allgemeine Tipps für das Entwerfen von Prompts"
href="/introduction/tips"
/>
<Card
icon={<FilesIcon />}
title="Beispiel für Prompts"
href="/introduction/examples"
/>
</Cards>
<ContentFileNames section="introduction" lang="de"/>

@ -2,6 +2,7 @@
import {Cards, Card} from 'nextra-theme-docs'
import { CardsIcon, OneIcon, WarningIcon, FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently apply and build with large language models (LLMs) for a wide variety of applications and use cases.
@ -11,31 +12,4 @@ This comprehensive guide covers the theory and practical aspects of prompt engin
All examples are tested with `gpt-3.5-turbo` using the [OpenAI's Playground](https://platform.openai.com/playground) unless otherwise specified. The model uses the default configurations, i.e., `temperature=1` and `top_p=1`. The prompts should also work with other models that have similar capabilities as `gpt-3.5-turbo` but the model responses may vary.
<Cards>
<Card
icon={<FilesIcon />}
title="LLM Settings"
href="/introduction/settings"
/>
<Card
icon={<FilesIcon />}
title="Basics of Prompting"
href="/introduction/basics"
/>
<Card
icon={<FilesIcon />}
title="Prompt Elements"
href="/introduction/elements"
/>
<Card
icon={<FilesIcon />}
title="General Tips for Designing Prompts"
href="/introduction/tips"
/>
<Card
icon={<FilesIcon />}
title="Examples of Prompts"
href="/introduction/examples"
/>
</Cards>
<ContentFileNames section="introduction" lang="en"/>

@ -1,7 +1,12 @@
# Introducción
import ContentFileNames from 'components/ContentFileNames'
La ingeniería de prompt es una disciplina relativamente nueva para el desarrollo y la optimización de prompts para utilizar eficientemente modelos de lenguaje (ML) en una amplia variedad de aplicaciones y temas de investigación. Las habilidades de ingeniería de prompt ayudan a comprender mejor las capacidades y limitaciones de los grandes modelos de lenguaje (LLM). Los investigadores utilizan la ingeniería de prompt para mejorar la capacidad de los LLM en una amplia gama de tareas comunes y complejas, como responder preguntas y razonamiento aritmético. Los desarrolladores utilizan la ingeniería de prompt para diseñar técnicas de prompt robustas y efectivas que interactúen con los LLM y otras herramientas.
Esta guía cubre los conceptos básicos de los prompts para proporcionar una idea general de cómo utilizar los prompts para interactuar e instruir a los grandes modelos de lenguaje (LLM).
Todos los ejemplos se han probado con `text-davinci-003` (usando el playground de OpenAI) a menos que se especifique lo contrario. Se utilizan las configuraciones predeterminadas, es decir, `temperature=0.7` y `top-p=1`.
<ContentFileNames section="introduction" lang="es"/>

@ -1,8 +1,13 @@
# Johdanto
import ContentFileNames from 'components/ContentFileNames'
Kehotesuunnittelu on suhteellisen uusi tieteenala, joka keskittyy kehotteiden kehittämiseen ja optimointiin. Sen avulla kielimalleja (Language Model, LM) voidaan käyttää tehokkaasti monenlaisissa sovelluksissa ja tutkimusaiheissa. Kehotesuunnittelun taidot auttavat ymmärtämään suurten kielimallien (Large Language Model, LLM) kykyjä ja rajoituksia paremmin. Tutkijat käyttävät kehotesuunnittelua parantaakseen LLM:ien kyvykkyyksiä erilaisissa tehtävissä, joista kysymyksiin vastaaminen ja aritmeettinen päättely ovat hyviä esimerkkejä. Kehittäjät käyttävät kehotesuunnittelua kestävien ja tehokkaiden kehotetekniikoiden kehittämiseen, jotka hyödyntävät LLM:ien potentiaalia optimaalisella tavalla.
Tämä opas käsittelee kehotteiden perusteita ja antaa yleiskuvan siitä, kuinka kehotteita voidaan käyttää vuorovaikutuksessa ja ohjeistuksessa LLM:ien kanssa.
Kaikki esimerkit on testattu `text-davinci-003` -mallilla käyttäen [OpenAI:n testiympäristöä](https://platform.openai.com/playground) ellei toisin mainita. Malli käyttää oletusasetuksia, eli `temperature=0.7` ja `top-p=1`.
<ContentFileNames section="introduction" lang="fi"/>

@ -1,7 +1,11 @@
# Introduction
import ContentFileNames from 'components/ContentFileNames'
Prompt engineering est une discipline relativement nouvelle visant à développer et à optimiser des prompts pour utiliser efficacement des modèles de langage (LMs) dans une grande variété d'applications et de sujets de recherche. Les compétences en prompt engineering aident à mieux comprendre les capacités et les limitations des grands modèles de langage (LLMs). Les chercheurs utilisent le prompt engineering pour améliorer la capacité des LLMs sur une large gamme de tâches courantes et complexes, telles que la réponse à des questions et le raisonnement arithmétique. Les développeurs utilisent également le prompt engineering pour concevoir des techniques de promptage robustes et efficaces qui interagissent avec les LLMs et d'autres outils.
Ce guide couvre les bases des prompts pour fournir une idée approximative de comment utiliser les prompts pour interagir et instruire les grands modèles de langage (LLMs).
Tous les exemples ont été testés avec text-davinci-003 (en utilisant le playground d'OpenAI), sauf indication contraire. Ils utilisent les configurations par défaut, c'est-à-dire temperature=0.7 et top-p=1.
Tous les exemples ont été testés avec text-davinci-003 (en utilisant le playground d'OpenAI), sauf indication contraire. Ils utilisent les configurations par défaut, c'est-à-dire temperature=0.7 et top-p=1.
<ContentFileNames section="introduction" lang="fr"/>

@ -1,5 +1,8 @@
# Introduzione
import ContentFileNames from 'components/ContentFileNames'
Il prompt engineering - ingegneria dei prompt - è una disciplina relativamente nuova per lo sviluppo e l'ottimizzazione dei prompt per utilizzare in modo efficiente i modelli linguistici (LM) per un'ampia varietà di applicazioni e argomenti di ricerca.
Le competenze di prompt engineering aiutano a comprendere meglio le capacità e i limiti dei modelli di linguaggio di grandi dimensioni (LLM). I ricercatori utilizzano il prompt engineering per migliorare la capacità degli LLM su un'ampia gamma di attività comuni e complesse come la risposta alle domande e il ragionamento aritmetico. Gli sviluppatori utilizzano il prompt engineering per progettare tecniche di prompt robuste ed efficaci che si interfacciano con LLM e altri strumenti.
@ -7,3 +10,6 @@ Le competenze di prompt engineering aiutano a comprendere meglio le capacità e
Questa guida copre le nozioni di base dei prompt per fornire un'idea approssimativa di come utilizzare i prompt per interagire e istruire modelli di linguaggi di grandi dimensioni (LLM).
Tutti gli esempi sono testati con `gpt-3.5-turbo` utilizzando il playground di OpenAI se non diversamente specificato. Utilizza le configurazioni predefinite, ovvero `temperature=1` e `top_p=1`. Gli esempi dovrebbero funzionare anche con altri modelli che hanno capacità simili a quelle di `gpt-3.5-turbo` ma le loro risposte potrebbero essere diverse.
<ContentFileNames section="introduction" lang="it"/>

@ -1,7 +1,12 @@
# はじめに
import ContentFileNames from 'components/ContentFileNames'
プロンプトエンジニアリングは、言語モデルLMを効率的に使用するためのプロンプトの開発と最適化のための比較的新しい学問分野です。プロンプトエンジニアリングのスキルは、大規模な言語モデルLLMの能力と限界をより良く理解するのに役立ちます。研究者は、プロンプトエンジニアリングを使用して、質問応答や算術推論などの一般的で複雑なタスクの幅広い範囲でLLMの能力を向上させます。開発者は、プロンプトエンジニアリングを使用して、LLMやその他のツールとインターフェースする堅牢で効果的なプロンプティング技術を設計します。
このガイドでは、プロンプトの基本をカバーし、大規模な言語モデルLLMとやり取りして指示する方法の概要を提供します。
すべての例は、OpenAIのプレイグラウンドを使用した `text-davinci-003` でテストされています。デフォルトの設定、すなわち `temperature = 0.7` および `top-p = 1` を使用しています。
すべての例は、OpenAIのプレイグラウンドを使用した `text-davinci-003` でテストされています。デフォルトの設定、すなわち `temperature = 0.7` および `top-p = 1` を使用しています。
<ContentFileNames section="introduction" lang="jp"/>

@ -1,7 +1,12 @@
# Introduction
import ContentFileNames from 'components/ContentFileNames'
프롬프트 엔지니어링은 다양한 어플리케이션과 연구 주제에 언어 모델(LMs)을 효율적으로 사용할 수 있도록 프롬프트를 개발하고 최적화하는 비교적 새로운 분야입니다. 프롬프트 엔지니어링 기술은 대규모 언어 모델(LLMs)의 기능과 한계를 더 잘 이해하는 데 도움이 됩니다. 연구자들은 프롬프트 엔지니어링을 사용하여 질문 답변 및 산술 추론과 같은 일반적이고 복잡한 다양한 작업에서 LLMs의 역량을 향상시킵니다. 개발자는 프롬프트 엔지니어링을 사용하여 LLMs 및 기타 도구와 인터페이스하는 강력하고 효과적인 프롬프트 기술을 설계합니다.
이 가이드는 프롬프트의 기본 사항을 다루며 프롬프트를 사용하여 대규모 언어 모델(LLMs)과 상호 작용하고 지시하는 방법에 대한 개략적인 아이디어를 제공합니다.
모든 예제는 달리 명시되지 않는 한 `text-davinci-003`(OpenAI의 플레이그라운드 사용)으로 테스트되었습니다. 기본 구성, 즉 `temperature=0.7` 및 `top-p=1`을 사용합니다.
모든 예제는 달리 명시되지 않는 한 `text-davinci-003`(OpenAI의 플레이그라운드 사용)으로 테스트되었습니다. 기본 구성, 즉 `temperature=0.7` 및 `top-p=1`을 사용합니다.
<ContentFileNames section="introduction" lang="kr"/>

@ -1,7 +1,12 @@
# Introdução
import ContentFileNames from 'components/ContentFileNames'
A engenharia de prompts é uma disciplina relativamente nova para desenvolver e otimizar prompts para usar eficientemente modelos de linguagem (LMs) para uma ampla variedade de aplicativos e tópicos de pesquisa. As habilidades imediatas de engenharia ajudam a entender melhor os recursos e as limitações dos modelos de linguagem grandes (LLMs). Os pesquisadores usam a engenharia de prompt para melhorar a capacidade dos LLMs em uma ampla gama de tarefas comuns e complexas, como resposta a perguntas e raciocínio aritmético. Os desenvolvedores usam engenharia de prompt para projetar técnicas de prompt robustas e eficazes que fazem interface com LLMs e outras ferramentas.
Este guia aborda os fundamentos dos prompts para fornecer uma ideia aproximada de como utiliza-los para interagir e instruir modelos de linguagem grandes (LLMs).
Todos os exemplos são testados com `text-davinci-003` (usando o playground do OpenAI), a menos que especificado de outra forma. Ele usa as configurações padrão, ou seja, `temperatura=0.7` e `top-p=1`.
Todos os exemplos são testados com `text-davinci-003` (usando o playground do OpenAI), a menos que especificado de outra forma. Ele usa as configurações padrão, ou seja, `temperatura=0.7` e `top-p=1`.
<ContentFileNames section="introduction" lang="pt"/>

@ -1,7 +1,12 @@
# Введение
import ContentFileNames from 'components/ContentFileNames'
Промпт-инжиниринг - это относительно новая дисциплина разработки и оптимизации промптов для эффективного использования языковых моделей (LM) в широком спектре приложений и исследовательских тем. Навыки промпт-инжиниринга помогают лучше понять возможности и ограничения больших языковых моделей (LLM). Исследователи используют промпт-инжиниринг для улучшения возможностей LLM на широком спектре общих и сложных задач, таких как вопросно-ответная система и арифметическое рассуждение. Разработчики используют промпт-инжиниринг для разработки надежных и эффективных методов промптинга, взаимодействующих с LLM и другими инструментами.
Это руководство охватывает основы промптов, чтобы дать общее представление о том, как использовать промпты для взаимодействия и командования LLM.
Все примеры протестированы с использованием `text-davinci-003` на [площадке OpenAI](https://platform.openai.com/playground), если не указано иное. Модель использует конфигурации по умолчанию, т.е. `temperature=0.7` и `top-p=1`.
Все примеры протестированы с использованием `text-davinci-003` на [площадке OpenAI](https://platform.openai.com/playground), если не указано иное. Модель использует конфигурации по умолчанию, т.е. `temperature=0.7` и `top-p=1`.
<ContentFileNames section="introduction" lang="ru"/>

@ -1,7 +1,12 @@
# Giriş
import ContentFileNames from 'components/ContentFileNames'
İstem mühendisliği, dil modellerini (LM'ler) çeşitli uygulamalar ve araştırma konuları için verimli bir şekilde kullanmak üzere istemleri geliştirme ve optimize etme konusunda nispeten yeni bir disiplindir. İstem mühendisliği becerileri, büyük dil modellerinin (LLM'ler) yeteneklerini ve sınırlamalarını daha iyi anlamaya yardımcı olur. Araştırmacılar, istem mühendisliğini, soru cevaplama ve aritmetik akıl yürütme gibi çeşitli ortak ve karmaşık görevlerde LLM'lerin kapasitesini artırmak için kullanırlar. Geliştiriciler, LLM'ler ve diğer araçlarla arayüz sağlayan sağlam ve etkili istem teknikleri tasarlamak için istem mühendisliğini kullanır.
Bu kılavuz, LLM'leri ile etkileşim kurmak ve yönlendirmek için istemleri nasıl kullanacağınıza dair genel bir fikir vermek üzere istemlerin temellerini kapsar.
Tüm örnekler, aksi belirtilmedikçe `text-davinci-003` kullanılarak [OpenAI's playground](https://platform.openai.com/playground) üzerinde test edilmiştir. Model, varsayılan yapılandırmaları kullanır, yani `temperature=0.7` ve `top-p=1`.
Tüm örnekler, aksi belirtilmedikçe `text-davinci-003` kullanılarak [OpenAI's playground](https://platform.openai.com/playground) üzerinde test edilmiştir. Model, varsayılan yapılandırmaları kullanır, yani `temperature=0.7` ve `top-p=1`.
<ContentFileNames section="introduction" lang="tr"/>

@ -1,7 +1,12 @@
# 提示工程简介
import ContentFileNames from 'components/ContentFileNames'
提示工程是一个较新的学科应用于开发和优化提示词Prompt帮助用户有效地将语言模型用于各种应用场景和研究领域。掌握了提示工程相关技能将有助于用户更好地了解大型语言模型的能力和局限性。研究人员可利用提示工程来提高大语言模型处理复杂任务场景的能力如问答和算术推理能力。开发人员可通过提示工程设计和研发出强大的技术实现和大语言模型或其他生态工具的高效接轨。
本指南介绍了提示词相关的基础知识,帮助用户了解如何通过提示词和大语言模型进行交互并提供指导建议。
除非特别说明,本指南默认所有示例都是基于 OpenAI 的大语言模型 `text-davinci-003` 进行测试,并且使用该模型的默认配置,如 `temperature=0.7` 和 `top-p=1` 等。
<ContentFileNames section="introduction" lang="zh"/>

@ -1,9 +1,9 @@
# Models
# Model Prompting Guides
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
En aquesta secció, tractarem alguns dels models de llenguatge més recents i com apliquen amb èxit les tècniques d'enginyeria de prompts més avançades i actuals. A més, cobrim les capacitats d'aquests models en una sèrie de tasques i configuracions de prompts, com ara sol·licituds amb poques mostres (few-shot prompting), sol·licituds sense mostres (zero-shot prompting) i sol·licituds en cadena de pensament (chain-of-thought prompting). Entendre aquestes capacitats és important per comprendre les limitacions d'aquests models i com utilitzar-los de manera efectiva.
<Callout emoji="⚠️">
Aquesta secció està en plena fase de desenvolupament.
</Callout>
<ContentFileNames section="models" lang="ca"/>

@ -3,28 +3,8 @@
import { Callout } from 'nextra-theme-docs';
import { Cards, Card } from 'nextra-theme-docs';
import { FilesIcon } from 'components/icons';
import ContentFileNames from 'components/ContentFileNames'
In diesem Abschnitt werden wir einige der neuesten Sprachmodelle behandeln und wie sie die neuesten und fortschrittlichsten Techniken im Prompting erfolgreich anwenden. Zusätzlich gehen wir auf die Fähigkeiten dieser Modelle bei einer Reihe von Aufgaben und Prompting-Setups ein, wie etwa Few-Shot Prompting, Zero-Shot Prompting und Chain-of-Thought Prompting. Das Verständnis dieser Fähigkeiten ist wichtig, um die Grenzen dieser Modelle zu verstehen und wie man sie effektiv einsetzt.
<Cards>
<Card icon={<FilesIcon />} title="Flan" href="/models/flan" />
<Card icon={<FilesIcon />} title="ChatGPT" href="/models/chatgpt" />
<Card icon={<FilesIcon />} title="LLaMA" href="/models/llama" />
<Card icon={<FilesIcon />} title="GPT-4" href="/models/gpt-4" />
<Card
icon={<FilesIcon />}
title="Mistral 7B"
href="/models/mistral-7b"
/>
<Card icon={<FilesIcon />} title="Gemini" href="/models/gemini" />
<Card
icon={<FilesIcon />}
title="Phi-2"
href="/models/phi-2"
/>
<Card
icon={<FilesIcon />}
title="LLM-Sammlung"
href="/models/collection"
/>
</Cards>
<ContentFileNames section="models" lang="de"/>

@ -3,49 +3,8 @@
import { Callout } from 'nextra-theme-docs'
import {Cards, Card} from 'nextra-theme-docs'
import {FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
In this section, we will cover some of the recent language models and how they successfully apply the latest and most advanced prompting engineering techniques. In addition, we cover capabilities of these models on a range of tasks and prompting setups like few-shot prompting, zero-shot prompting, and chain-of-thought prompting. Understanding these capabilities are important to understand the limitations of these models and how to use them effectively.
<Cards>
<Card
icon={<FilesIcon />}
title="Flan"
href="/models/flan"
/>
<Card
icon={<FilesIcon />}
title="ChatGPT"
href="/models/chatgpt"
/>
<Card
icon={<FilesIcon />}
title="Llama"
href="/models/llama"
/>
<Card
icon={<FilesIcon />}
title="GPT-4"
href="/models/gpt-4"
/>
<Card
icon={<FilesIcon />}
title="Mistral 7B"
href="/models/mistral-7b"
/>
<Card
icon={<FilesIcon />}
title="Gemini"
href="/models/gemini"
/>
<Card
icon={<FilesIcon />}
title="Phi-2"
href="/models/phi-2"
/>
<Card
icon={<FilesIcon />}
title="LLM Collection"
href="/models/collection"
/>
</Cards>
<ContentFileNames section="models" lang="en"/>

@ -1,9 +1,9 @@
# Modelos
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
En esta sección, cubriremos algunos de los modelos de lenguaje más recientes y cómo aplican con éxito las últimas y más avanzadas técnicas de ingeniería de generación de texto. Además, abarcamos las capacidades de estos modelos en una variedad de tareas y configuraciones de generación de texto, como la generación de texto con pocos ejemplos, la generación de texto sin ejemplos y la generación de texto de encadenamiento de pensamiento. Comprender estas capacidades es importante para entender las limitaciones de estos modelos y cómo utilizarlos de manera efectiva.
<Callout emoji="⚠️">
Esta sección está en pleno desarrollo.
</Callout>
<ContentFileNames section="models" lang="es"/>

@ -1,9 +1,8 @@
# Mallit
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Tässä osiossa käsittelemme joitakin viimeaikaisia kielimalleja ja kuinka ne soveltavat menestyksekkäästi uusimpia ja edistyneimpiä kehotteita. Lisäksi tarkastelemme näiden mallien suorituskykyä monenlaisissa tehtävissä ja ohjausasetuksissa, kuten vähäisessä ohjauksessa, nollaohjauksessa ja ajatusketjuohjauksessa. Näiden kykyjen ymmärtäminen on tärkeää tunnistaaksemme mallien rajoitukset ja kuinka niitä voidaan käyttää tehokkaasti.
<Callout emoji="⚠️">
Tämä osa sivustoa kehittyy jatkuvasti.
</Callout>
<ContentFileNames section="models" lang="fi"/>

@ -1,9 +1,10 @@
# Models
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Dans cette section, nous aborderons certains des modèles de langage récents et comment ils appliquent avec succès les techniques les plus avancées de prompting engineering.Nous couvrirons également les capacités de ces modèles sur une gamme de tâches et de configurations de promptage, telles que le promptage à quelques exemples, le promptage à zéro exemple et le promptage en chaîne de pensées. Comprendre ces capacités est important pour comprendre les limites de ces modèles et comment les utiliser efficacement.
<Callout emoji="⚠️">
Cette section est en plein développement.
</Callout>
<ContentFileNames section="models" lang="fr"/>

@ -1,9 +1,9 @@
# Modelli
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
In questa sezione, verranno illustrati alcuni dei recenti modelli linguistici e il modo in cui essi applicano con successo le più recenti e avanzate tecniche di prompting. Inoltre, vengono descritte le capacità di questi modelli su una serie di compiti e configurazioni di prompt, come il prompt a pochi colpi, il prompt a zero colpi e il prompt a catena di pensieri. La comprensione di queste capacità è importante per capire i limiti di questi modelli e come utilizzarli in modo efficace.
<Callout emoji="⚠️">
Questa sezione è in fase di forte sviluppo.
</Callout>
<ContentFileNames section="models" lang="it"/>

@ -1,9 +1,9 @@
# モデル
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
このセクションでは、最近の言語モデルを取り上げ、それらがどのように最新かつ最も高度なプロンプト工学技術をうまく適用しているかを説明します。さらに、これらの言語モデルの能力を、様々なタスクやプロンプトの設定、例えばfew-shotプロンプト、zero-shotプロンプト、chain-of-thoughtプロンプトについて説明します。これらの機能を理解することは、これらのモデルの限界を理解し、効果的に使用する方法として重要です。
<Callout emoji="⚠️">
このセクションは、現在開発が進んでいます。
</Callout>
<ContentFileNames section="models" lang="jp"/>

@ -1,9 +1,10 @@
# Models
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
이 장에서는 몇 가지 최신 언어 모델과 이 모델들이 최신의 첨단 프롬프트 엔지니어링 기법을 효과적으로 적용하는 방법을 다룹니다. 또한 few-shot prompting, zero-shot prompting, and chain-of-thought prompting과 같은 다양한 작업 및 프롬프트 설정에 대한 이러한 모델의 기능에 대해서도 다룹니다. 이러한 기능을 이해하는 것은 모델들의 한계를 이해하고 효과적으로 사용하는데 중요합니다.
<Callout emoji="⚠️">
This section is under heavy development.
</Callout>
<ContentFileNames section="models" lang="fr"/>

@ -1,9 +1,9 @@
# Modelos
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Nesta seção, abordaremos alguns dos modelos de linguagem recentes e como eles aplicam com êxito as técnicas de engenharia de solicitação mais recentes e avançadas. Além disso, abordamos os recursos desses modelos em uma variedade de tarefas e configurações de solicitação, como solicitação de poucos disparos, solicitação de disparo zero e solicitação de cadeia de pensamento. Entender esses recursos é importante para entender as limitações desses modelos e como usá-los de forma eficaz.
<Callout emoji="⚠️">
Esta seção está em intenso desenvolvimento.
</Callout>
<ContentFileNames section="models" lang="pt"/>

@ -1,9 +1,9 @@
# Modeller
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Bu bölümde, en yeni dil modellerinden bazılarını ve bunların en yeni ve en gelişmiş yönlendirme mühendisliği tekniklerini nasıl başarıyla uyguladıklarını ele alacağız. Ek olarak, bu modellerin bir dizi görevdeki yeteneklerini ve az örnekli yönlendirme, sıfır örnekli yönlendirme ve düşünce zinciri yönlendirmesi gibi komut istemi kurulumlarını ele alıyoruz. Bu yetenekleri anlamak, bu modellerin sınırlamalarını ve bunların nasıl etkili bir şekilde kullanılacağını anlamak için önemlidir.
<Callout emoji="⚠️">
Bu bölüm yoğun bir geliştirme aşamasındadır.
</Callout>
<ContentFileNames section="models" lang="tr"/>

@ -1,9 +1,9 @@
# 模型
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
在本节中,我们将介绍一些最近的语言模型以及它们如何成功地应用最新和最先进的提示工程技术。此外,我们还将介绍这些模型在各种任务和提示设置(如少样本提示、零样本提示和思维链提示)中的能力。了解这些能力对于理解这些模型的局限性以及如何有效地使用它们非常重要。
<Callout emoji="⚠️">
本节正在大力开发中。
</Callout>
<ContentFileNames section="models" lang="zh"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="ca"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="de"/>

@ -2,17 +2,11 @@
import {Cards, Card} from 'nextra-theme-docs'
import {FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
In this section, we regularly highlight miscellaneous and interesting research findings about how to better work with large language models (LLMs). It include new tips, insights and developments around important LLM research areas such as scaling, agents, efficiency, hallucination, architectures, prompt injection, and much more.
LLM research and AI research in general is moving fast so we hope that this resource can help both researchers and developers stay ahead of important developments. We also welcome contributions to this section if you would like to highlight an exciting finding about your research or experiments.
<Cards>
<Card
icon={<FilesIcon />}
title="Trustworthiness in LLMs"
href="/research/trustworthiness-in-llms"
/>
</Cards>
<ContentFileNames section="research" lang="en"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="es"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="fi"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="fr"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="it"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="jp"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="kr"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="pt"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="ru"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="tr"/>

@ -1,3 +1,7 @@
# LLM Research Findings
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
import ContentFileNames from 'components/ContentFileNames'
This page needs a translation! Feel free to contribute a translation by clicking the `Edit this page` button on the right.
<ContentFileNames section="research" lang="zh"/>

@ -58,6 +58,4 @@ Code: https://github.com/HowieHwong/TrustLLM
## References
Image Source / Paper: [TrustLLM: Trustworthiness in Large Language Models](https://arxiv.org/abs/2401.05561) (10 Jan 2024)

@ -1,11 +1,10 @@
# Riscs i Mal ús
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Ja hem vist com de efectives poden ser els prompts ben dissenyats per a diverses tasques utilitzant tècniques com l'aprenentatge amb poques mostres i l'encadenament de pensaments. A mesura que pensem en construir aplicacions reals basades en LLMs, esdevé crucial reflexionar sobre els mal ús, riscs i pràctiques de seguretat relacionats amb els models de llenguatge.
Aquesta secció se centra en destacar alguns dels riscs i mal ús dels LLMs mitjançant tècniques com injeccions de prompts. També destaca comportaments perjudicials i com potencialment mitigar-los mitjançant tècniques de prompts efectives. Altres temes d'interès inclouen generalitzabilitat, calibratge, biaixos, biaixos socials i factualitat, per esmentar-ne alguns.
<Callout emoji="⚠️">
Aquesta secció està sota un intens desenvolupament.
</Callout>
<ContentFileNames section="risks" lang="ca"/>

@ -3,25 +3,11 @@
import { Callout } from 'nextra-theme-docs';
import { Cards, Card } from 'nextra-theme-docs';
import { FilesIcon } from 'components/icons';
import ContentFileNames from 'components/ContentFileNames'
Gut formulierte Prompts können zu effektivem Einsatz von LLMs für verschiedene Aufgaben unter Verwendung von Techniken wie Few-Shot-Learning und Chain-of-Thought-Prompts führen. Sobald Sie darüber nachdenken, reale Anwendungen auf Basis von LLMs zu entwickeln, wird es auch entscheidend, über Missbräuche, Risiken und Sicherheitspraktiken im Zusammenhang mit Sprachmodellen nachzudenken.
Dieser Abschnitt konzentriert sich darauf, einige der Risiken und Missbräuche von LLMs mittels Techniken wie Prompt-Injektionen hervorzuheben. Es beleuchtet auch schädliche Verhaltensweisen und wie diese möglicherweise durch effektive Prompting-Techniken und Tools wie Moderations-APIs gemildert werden können. Andere interessante Themen umfassen Allgemeingültigkeit, Kalibrierung, Voreingenommenheiten, soziale Verzerrungen und Faktentreue, um nur einige zu nennen.
<Cards>
<Card
icon={<FilesIcon />}
title="Adversariales Prompting"
href="/risks/adversarial"
/>
<Card
icon={<FilesIcon />}
title="Faktentreue"
href="/risks/factuality"
/>
<Card
icon={<FilesIcon />}
title="Verzerrungen (biases)"
href="/risks/biases"
/>
</Cards>
<ContentFileNames section="risks" lang="de"/>

@ -3,26 +3,11 @@
import { Callout } from 'nextra-theme-docs'
import {Cards, Card} from 'nextra-theme-docs'
import {FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
Well-crafted prompts can lead to effective used of LLMs for various tasks using techniques like few-shot learning and chain-of-thought prompting. As you think about building real-world applications on top of LLMs, it also becomes crucial to think about the misuses, risks, and safety practices involved with language models.
This section focuses on highlighting some of the risks and misuses of LLMs via techniques like prompt injections. It also highlights harmful behaviors and how to potentially mitigate them via effective prompting techniques and tools like moderation APIs. Other topics of interest include generalizability, calibration, biases, social biases, and factuality to name a few.
<Cards>
<Card
icon={<FilesIcon />}
title="Adversarial Prompting"
href="/risks/adversarial"
/>
<Card
icon={<FilesIcon />}
title="Factuality"
href="/risks/factuality"
/>
<Card
icon={<FilesIcon />}
title="Biases"
href="/risks/biases"
/>
</Cards>
<ContentFileNames section="risks" lang="en"/>

@ -1,11 +1,12 @@
# Riesgos y Malos Usos
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Ya hemos visto lo efectivos que pueden ser los prompts bien elaborados para varias tareas utilizando técnicas como el aprendizaje de pocos ejemplos (few-shot learning) y el prompting encadenado (chain-of-thought prompting). A medida que pensamos en construir aplicaciones del mundo real sobre modelos de lenguaje de gran escala (LLMs, por sus siglas en inglés), se vuelve crucial pensar en los abusos, riesgos y prácticas de seguridad involucradas con los modelos de lenguaje.
Esta sección se enfoca en destacar algunos de los riesgos y abusos de los LLMs a través de técnicas como la inyección de prompts. También destaca comportamientos dañinos y cómo mitigarlos potencialmente mediante técnicas de prompting efectivas. Otros temas de interés incluyen la generalización, la calibración, los sesgos, los sesgos sociales y la veracidad, por nombrar algunos.
<Callout emoji="⚠️">
Esta sección está en pleno desarrollo.
</Callout>
<ContentFileNames section="risks" lang="es"/>

@ -1,11 +1,11 @@
# Riskit ja väärinkäytökset
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Olemme nähneet kuinka tehokkaita hyvin muotoillut kehotteet voivat olla erilaisten tehtävien suorittamiseen tekniikoilla, kuten vähäisen ohjauksen kehottaminen ja ajatusketjuohjaus. Kun mietimme sovellusten rakentamista LLM:ien päälle, on tärkeää pohtia väärinkäytöksiä, riskejä ja turvallisuuskäytäntöjä, jotka liittyvät kielimalleihin.
Tämä osio keskittyy korostamaan joitakin LLM:ien riskien ja väärinkäytösten tekniikoita, kuten kehoteinjektiot. Osio tuo esiin myös haitalliset käyttäytymismallit ja sen, kuinka niitä voidaan mahdollisesti lieventää tehokkailla kehotesuunnittelutekniikoilla. Muita kiinnostavia aiheita ovat yleistettävyys, kalibrointi, vinoumat, sosiaaliset vinoumat ja faktuaalisuus, vain muutamia mainitakseni.
<Callout emoji="⚠️">
Tämä osa sivustoa kehittyy jatkuvasti.
</Callout>
<ContentFileNames section="risks" lang="fi"/>

@ -1,11 +1,12 @@
# Risks & Misuses
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Nous avons déjà vu à quel point des prompts bien conçus peuvent être efficaces pour diverses tâches en utilisant des techniques telles que l'apprentissage à quelques exemples et la stimulation de la chaîne de pensée. En envisageant de construire des applications concrètes sur la base de modèles de langage de grande envergure, il est crucial de réfléchir aux utilisations abusives, aux risques et aux pratiques de sécurité liées aux modèles de langage.
Cette section met l'accent sur la mise en évidence de certains des risques et abus associés aux LLMs via des techniques telles que les injections de prompts. Elle met également en évidence les comportements préjudiciables et la façon de les atténuer potentiellement grâce à des techniques de prompt efficaces. D'autres sujets d'intérêt comprennent la généralisabilité, l'étalonnage, les biais, les biais sociaux et la factualité, pour n'en nommer que quelques-uns.
<Callout emoji="⚠️">
Cette section est en plein développement.
</Callout>
<ContentFileNames section="risks" lang="fr"/>

@ -1,11 +1,11 @@
# Rischi e Abusi
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Abbiamo già visto quanto possano essere efficaci i prompt ben fatti per vari compiti, utilizzando tecniche come l'apprendimento a pochi colpi e il prompt a catena di pensieri. Nel momento in cui pensiamo di costruire applicazioni reali sulla base dei LLM, diventa fondamentale riflettere sugli usi impropri, sui rischi e sulle pratiche di sicurezza che i modelli linguistici comportano.
Questa sezione si concentra sull'evidenziazione di alcuni rischi e usi impropri degli LLM attraverso tecniche come le iniezioni di prompt. Vengono inoltre evidenziati i comportamenti dannosi e le modalità per mitigarli potenzialmente attraverso tecniche di prompting efficaci. Altri argomenti di interesse sono la generalizzabilità, la calibrazione, i pregiudizi, i pregiudizi sociali e la fattualità, per citarne alcuni.
<Callout emoji="⚠️">
Questa sezione è in fase di forte sviluppo.
</Callout>
<ContentFileNames section="risks" lang="it"/>

@ -1,11 +1,11 @@
# リスクと誤用
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
私たちは、few-shot学習やchain-of-thoughtプロンプトのようなテクニックを使って、うまく作られたプロンプトがさまざまなタスクでいかに効果的であるかをすでに見てきました。LLMの上に実世界のアプリケーションを構築することを考えると、言語モデルの誤用、リスク、安全対策について考えることが非常に重要になります。
このセクションでは、プロンプトインジェクションのような手法によるLLMのリスクと誤用に焦点を当てます。また、有害な行動と、効果的なプロンプト技術によってそれを軽減する方法についても言及します。その他、一般化可能性、キャリブレーション、バイアス、社会的バイアス、事実性など、興味のあるトピックをいくつか挙げていきます。
<Callout emoji="⚠️">
このセクションは、現在開発が進んでいます。
</Callout>
<ContentFileNames section="risks" lang="jp"/>

@ -1,11 +1,11 @@
# Risks & Misuses
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
우리는 이미 잘 만들어진 프롬프트가 few-shot learning and chain-of-thought prompting과 같은 기법을 사용하여 다양한 작업에 얼마나 효과적인지 보았습니다. LLMs을 기반으로 실제 어플리케이션을 구축할 때 언어 모델과 관련된 오용, 위험 및 안전 관행에 대해 생각하는 것이 중요해졌습니다.
이 장에서는 프롬프트 삽입과 같은 기술을 통해 LLMs의 몇 가지 위험과 오용을 강조하는 데 중점을 둡니다. 또한 유해한 행동을 지적하고, 효과적인 프롬프트 기술을 통해 이를 잠재적으로 완화할 수 있는 방법을 강조합니다. 그 밖에도 일반화 가능성, 보정, 편향성, 사회적 편견, 사실성 등 다양한 주제를 다룹니다.
<Callout emoji="⚠️">
This section is under heavy development.
</Callout>
<ContentFileNames section="risks" lang="kr"/>

@ -1,11 +1,11 @@
# Riscos e usos indevidos
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Já vimos como os prompts bem elaborados podem ser eficazes para várias tarefas, usando técnicas como aprendizado de poucos tiros e prompts de cadeia de pensamento. À medida que pensamos em construir aplicativos do mundo real sobre LLMs, torna-se crucial pensar sobre os usos indevidos, riscos e práticas de segurança envolvidas com modelos de linguagem.
Esta seção se concentra em destacar alguns dos riscos e usos indevidos de LLMs por meio de técnicas como injeções de prompt. Ele também destaca comportamentos nocivos e como potencialmente mitigá-los por meio de técnicas de alerta eficazes. Outros tópicos de interesse incluem generalização, calibração, vieses, vieses sociais e factualidade, para citar alguns.
<Callout emoji="⚠️">
Esta seção está em intenso desenvolvimento.
</Callout>
<ContentFileNames section="risks" lang="pt"/>

@ -1,11 +1,11 @@
# Риски и неправильное использование
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Мы уже видели, насколько эффективным могут быть хорошо проработанные промпты для различных задач, используя такие техники, как обучение с малым количеством примеров и промптинг на основе цепочки мыслей. Когда мы думаем о создании приложений для реального мира на основе больших языковых моделей (LLM), становится важным задуматься о возможных рисках, неправильном использовании и практиках безопасности, связанных с языковыми моделями.
Этот раздел сосредоточен на выявлении некоторых рисков и неправильного использования LLM с помощью таких техник, как внедрение промптов. Он также обращает внимание на вредоносное поведение и потенциальные способы смягчения его с помощью эффективных техник промптинга. Другие интересующие темы включают обобщаемость, калибровку, смещения, социальные предубеждения и достоверность, чтобы назвать некоторые из них.
<Callout emoji="⚠️">
Этот раздел находится в активной разработке.
</Callout>
<ContentFileNames section="risks" lang="ru"/>

@ -1,11 +1,11 @@
# Riskler & Kötüye Kullanımlar
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
Az sayıda örnek öğrenme ve düşünce zinciri oluşturma teknikleri gibi teknikleri kullanarak çeşitli görevler için ne kadar etkili olabileceğini gördük. LLM'lerin üzerine gerçek dünya uygulamaları oluşturmayı düşünürken, dil modelleriyle ilgili kötüye kullanım, riskler ve güvenlik uygulamaları hakkında düşünmek hayati önem taşır.
Bu bölüm, istem enjeksiyonları gibi teknikler aracılığıyla LLM'lerin risklerini ve kötüye kullanımlarını vurgulamaya odaklanır. Ayrıca zararlı davranışları ve bu tür davranışları etkili istem teknikleri aracılığıyla nasıl hafifletebileceğinizi vurgular. İlgilendiğimiz diğer konular arasında genelleştirilebilirlik, kalibrasyon, önyargılar, sosyal önyargılar ve gerçeklik sayılabilir.
<Callout emoji="⚠️">
Bu bölüm yoğun bir geliştirme aşamasındadır.
</Callout>
<ContentFileNames section="risks" lang="tr"/>

@ -1,11 +1,11 @@
# 风险和误用
import { Callout } from 'nextra-theme-docs'
import ContentFileNames from 'components/ContentFileNames'
我们已经看到了如何使用few-shot学习和链式思考提示等技术来完成各种任务有效的精心制作的提示是多么的有效。当我们考虑在LLMs之上构建实际应用程序时思考与语言模型相关的误用、风险和安全实践变得至关重要。
本节重点介绍了通过提示注入等技术来突出LLMs的一些风险和误用。它还强调了有害行为以及如何通过有效的提示技术来潜在地减轻它们。其他感兴趣的主题包括泛化能力、校准、偏见、社会偏见和事实性等等。
<Callout emoji="⚠️">
本节正在积极开发中。
</Callout>
<ContentFileNames section="risks" lang="zh"/>

@ -1,5 +1,9 @@
# Tècniques de Prompts
import ContentFileNames from 'components/ContentFileNames'
Fins ara, hauria de ser evident que ajuda a millorar els prompts per obtenir millors resultats en diferents tasques. Aquesta és la idea principal darrere l'enginyeria de prompts.
Encara que els exemples bàsics eren divertits, en aquesta secció tractem tècniques d'enginyeria de prompts més avançades que ens permeten aconseguir tasques més complexes i interessants.
Encara que els exemples bàsics eren divertits, en aquesta secció tractem tècniques d'enginyeria de prompts més avançades que ens permeten aconseguir tasques més complexes i interessants.
<ContentFileNames section="techniques" lang="ca"/>

@ -8,85 +8,10 @@ import {
FilesIcon,
} from 'components/icons';
import ContentFileNames from 'components/ContentFileNames'
Das Prompt-Engineering hilft dabei, Prompts effektiv zu gestalten und zu verbessern, um bessere Ergebnisse bei verschiedenen Aufgaben mit LLMs zu erzielen.
Während die vorherigen grundlegenden Beispiele unterhaltsam waren, behandeln wir in diesem Abschnitt fortgeschrittenere Techniken des Prompt-Engineerings, die es uns ermöglichen, komplexere Aufgaben zu bewältigen und die Zuverlässigkeit und Leistung von LLMs zu verbessern.
<Cards>
<Card
icon={<FilesIcon />}
title="Zero-Shot Prompting"
href="/techniques/zeroshot"
/>
<Card
icon={<FilesIcon />}
title="Few-Shot Prompting"
href="/techniques/fewshot"
/>
<Card
icon={<FilesIcon />}
title="Chain-of-Thought Prompting"
href="/techniques/cot"
/>
<Card
icon={<FilesIcon />}
title="Selbstkonsistenz"
href="/techniques/consistency"
/>
<Card
icon={<FilesIcon />}
title="Generiertes Wissens-Prompting"
href="/techniques/knowledge"
/>
<Card
icon={<FilesIcon />}
title="Prompt-Verkettung"
href="/techniques/prompt_chaining"
/>
<Card
icon={<FilesIcon />}
title="Tree of Thoughts Prompting"
href="/techniques/tot"
/>
<Card
icon={<FilesIcon />}
title="Retrieval Augmented Generation"
href="/techniques/rag"
/>
<Card
icon={<FilesIcon />}
title="Automatic Reasoning and Tool-Use"
href="/techniques/art"
/>
<Card
icon={<FilesIcon />}
title="Automatic Prompt Engineer"
href="/techniques/ape"
/>
<Card
icon={<FilesIcon />}
title="Active-Prompt"
href="/techniques/activeprompt"
/>
<Card
icon={<FilesIcon />}
title="Direct Stimulus Prompting"
href="/techniques/dsp"
/>
<Card
icon={<FilesIcon />}
title="Program-Aided Language"
href="/techniques/pal"
/>
<Card icon={<FilesIcon />} title="ReAct" href="/techniques/react" />
<Card
icon={<FilesIcon />}
title="Multimodal CoT"
href="/techniques/multimodalcot"
/>
<Card
icon={<FilesIcon />}
title="Graph-Prompting"
href="/techniques/graph"
/>
</Cards>
<ContentFileNames section="techniques" lang="de"/>

@ -2,90 +2,10 @@
import {Cards, Card} from 'nextra-theme-docs'
import { CardsIcon, OneIcon, WarningIcon, FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
Prompt Engineering helps to effectively design and improve prompts to get better results on different tasks with LLMs.
While the previous basic examples were fun, in this section we cover more advanced prompting engineering techniques that allow us to achieve more complex tasks and improve reliability and performance of LLMs.
<Cards>
<Card
icon={<FilesIcon />}
title="Zero-Shot Prompting"
href="/techniques/zeroshot"
/>
<Card
icon={<FilesIcon />}
title="Few-Shot Prompting"
href="/techniques/fewshot"
/>
<Card
icon={<FilesIcon />}
title="Chain-of-Thought Prompting"
href="/techniques/cot"
/>
<Card
icon={<FilesIcon />}
title="Self-Consistency"
href="/techniques/consistency"
/>
<Card
icon={<FilesIcon />}
title="General Knowledge Prompting"
href="/techniques/knowledge"
/>
<Card
icon={<FilesIcon />}
title="Prompt Chaining"
href="/techniques/prompt_chaining"
/>
<Card
icon={<FilesIcon />}
title="Tree of Thoughts Prompting"
href="/techniques/tot"
/>
<Card
icon={<FilesIcon />}
title="Retrieval Augmented Generation"
href="/techniques/rag"
/>
<Card
icon={<FilesIcon />}
title="Automatic Reasoning and Tool-Use"
href="/techniques/art"
/>
<Card
icon={<FilesIcon />}
title="Automatic Prompt Engineer"
href="/techniques/ape"
/>
<Card
icon={<FilesIcon />}
title="Active-Prompt"
href="/techniques/activeprompt"
/>
<Card
icon={<FilesIcon />}
title="Direct Stimulus Prompting"
href="/techniques/dsp"
/>
<Card
icon={<FilesIcon />}
title="Program-Aided Language"
href="/techniques/pal"
/>
<Card
icon={<FilesIcon />}
title="ReAct"
href="/techniques/react"
/>
<Card
icon={<FilesIcon />}
title="Multimodal CoT"
href="/techniques/multimodalcot"
/>
<Card
icon={<FilesIcon />}
title="Graph Prompting"
href="/techniques/graph"
/>
</Cards>
<ContentFileNames section="techniques" lang="en"/>

@ -1,5 +1,9 @@
# Técnicas de prompting
import ContentFileNames from 'components/ContentFileNames'
A estas alturas, debería ser obvio que mejorar los prompts ayuda a obtener mejores resultados en diferentes tareas. Esa es la idea principal detrás de la ingeniería de prompts.
Si bien los ejemplos básicos fueron divertidos, en esta sección cubriremos técnicas más avanzadas de ingeniería de prompts que nos permiten lograr tareas más complejas e interesantes.
<ContentFileNames section="techniques" lang="es"/>

@ -1,5 +1,9 @@
# Kehotesuunnittelutekniikat
import ContentFileNames from 'components/ContentFileNames'
Tähän mennessä pitäisi olla selvää, että kehotteiden parantaminen auttaa saamaan parempia tuloksia eri tehtävissä. Se on koko kehotesuunnittelun idea.
Vaikka perusesimerkit olivat hauskoja, tässä osiossa käsittelemme edistyneempiä kehotesuunnittelutekniikoita, joiden avulla voimme suorittaa monimutkaisempia ja mielenkiintoisempia tehtäviä.
Vaikka perusesimerkit olivat hauskoja, tässä osiossa käsittelemme edistyneempiä kehotesuunnittelutekniikoita, joiden avulla voimme suorittaa monimutkaisempia ja mielenkiintoisempia tehtäviä.
<ContentFileNames section="techniques" lang="fi"/>

@ -1,5 +1,10 @@
# Prompting Techniques
import ContentFileNames from 'components/ContentFileNames'
À ce stade, il devrait être évident que l'amélioration des prompts contribue à obtenir de meilleurs résultats sur différentes tâches. C'est l'idée principale derrière l'ingénierie de prompts.
Bien que les exemples de base aient été amusants, dans cette section, nous abordons des techniques plus avancées d'ingénierie de prompts qui nous permettent d'accomplir des tâches plus complexes et intéressantes.
Bien que les exemples de base aient été amusants, dans cette section, nous abordons des techniques plus avancées d'ingénierie de prompts qui nous permettent d'accomplir des tâches plus complexes et intéressantes.
<ContentFileNames section="techniques" lang="fr"/>

@ -2,90 +2,10 @@
import {Cards, Card} from 'nextra-theme-docs'
import { CardsIcon, OneIcon, WarningIcon, FilesIcon} from 'components/icons'
import ContentFileNames from 'components/ContentFileNames'
L'ingegneria dei prompt aiuta a progettare e migliorare efficacemente i prompt per ottenere risultati migliori in diversi compiti con gli LLM.
Se gli esempi di base sono stati divertenti, in questa sezione tratteremo tecniche più avanzate di ingegneria dei prompt, che ci permettono di realizzare compiti più complessi e di migliorare l'affidabilità e le prestazioni degli LLM.
<Cards>
<Card
icon={<FilesIcon />}
title="Zero-Shot Prompting"
href="/techniques/zeroshot"
/>
<Card
icon={<FilesIcon />}
title="Few-Shot Prompting"
href="/techniques/fewshot"
/>
<Card
icon={<FilesIcon />}
title="Chain-of-Thought Prompting"
href="/techniques/cot"
/>
<Card
icon={<FilesIcon />}
title="Self-Consistency"
href="/techniques/consistency"
/>
<Card
icon={<FilesIcon />}
title="General Knowledge Prompting"
href="/techniques/knowledge"
/>
<Card
icon={<FilesIcon />}
title="Prompt Chaining"
href="/techniques/prompt_chaining"
/>
<Card
icon={<FilesIcon />}
title="Tree of Thoughts Prompting"
href="/techniques/tot"
/>
<Card
icon={<FilesIcon />}
title="Retrieval Augmented Generation"
href="/techniques/rag"
/>
<Card
icon={<FilesIcon />}
title="Automatic Reasoning and Tool-Use"
href="/techniques/art"
/>
<Card
icon={<FilesIcon />}
title="Automatic Prompt Engineer"
href="/techniques/ape"
/>
<Card
icon={<FilesIcon />}
title="Active-Prompt"
href="/techniques/activeprompt"
/>
<Card
icon={<FilesIcon />}
title="Direct Stimulus Prompting"
href="/techniques/dsp"
/>
<Card
icon={<FilesIcon />}
title="Program-Aided Language"
href="/techniques/pal"
/>
<Card
icon={<FilesIcon />}
title="ReAct"
href="/techniques/react"
/>
<Card
icon={<FilesIcon />}
title="Multimodal CoT"
href="/techniques/multimodalcot"
/>
<Card
icon={<FilesIcon />}
title="Graph Prompting"
href="/techniques/graph"
/>
</Cards>
<ContentFileNames section="techniques" lang="it"/>

@ -1,5 +1,9 @@
# プロンプトエンジニアリング技術
import ContentFileNames from 'components/ContentFileNames'
この時点で明らかになっているように、異なるタスクでより良い結果を得るために、プロンプトを改善することが役立つことがわかりました。これがプロンプトエンジニアリングのアイデア全体です。
基本的な例は楽しかったですが、このセクションでは、より高度なプロンプトエンジニアリング技術を紹介し、より複雑で興味深いタスクを達成することができます。
基本的な例は楽しかったですが、このセクションでは、より高度なプロンプトエンジニアリング技術を紹介し、より複雑で興味深いタスクを達成することができます。
<ContentFileNames section="techniques" lang="jp"/>

@ -1,5 +1,9 @@
# Prompting Techniques
import ContentFileNames from 'components/ContentFileNames'
이쯤 되면 프롬프트를 개선하여 다양한 작업에서 더 나은 결과를 얻는 것이 도움이 된다는 것이 분명해졌을 것입니다. 이것이 바로 프롬프트 엔지니어링의 기본 개념입니다.
기본적인 예제는 재미있었지만, 이 장에서는 더 복잡하고 흥미로운 작업을 수행할 수 있는 고급 프롬프트 엔지니어링 기법을 다룹니다.
기본적인 예제는 재미있었지만, 이 장에서는 더 복잡하고 흥미로운 작업을 수행할 수 있는 고급 프롬프트 엔지니어링 기법을 다룹니다.
<ContentFileNames section="techniques" lang="kr"/>

@ -1,5 +1,9 @@
# Técnicas de Prompting
import ContentFileNames from 'components/ContentFileNames'
A essa altura, deve ser óbvio que ajuda a melhorar os prompts para obter melhores resultados em diferentes tarefas. Essa é a ideia por trás da engenharia de prompt.
Embora os exemplos básicos tenham sido divertidos, nesta seção abordamos técnicas de engenharia de solicitação mais avançadas que nos permitem realizar tarefas mais complexas e interessantes.
Embora os exemplos básicos tenham sido divertidos, nesta seção abordamos técnicas de engenharia de solicitação mais avançadas que nos permitem realizar tarefas mais complexas e interessantes.
<ContentFileNames section="techniques" lang="pt"/>

@ -1,5 +1,10 @@
# Техники промптинга
import ContentFileNames from 'components/ContentFileNames'
На данном этапе уже становится очевидным, что улучшение формулировки запросов помогает достичь лучших результатов в различных задачах. Вот основная идея, стоящая за техниками промптинга.
Хотя базовые примеры были интересными, в этом разделе мы рассмотрим более продвинутые техники формулировки запросов, которые позволяют нам решать более сложные и интересные задачи.
Хотя базовые примеры были интересными, в этом разделе мы рассмотрим более продвинутые техники формулировки запросов, которые позволяют нам решать более сложные и интересные задачи.
<ContentFileNames section="techniques" lang="ru"/>

@ -1,5 +1,9 @@
# İstem Teknikleri
import ContentFileNames from 'components/ContentFileNames'
Bu aşamada, farklı görevlerde daha iyi sonuçlar elde etmek için istemleri geliştirmenin yardımcı olduğu açık olmalıdır. Bu, istem mühendisliğinin tüm fikrinin arkasındadır.
Temel örnekler eğlenceli olsada, bu bölümde daha karmaşık ve ilginç görevler gerçekleştirmemize olanak sağlayan daha gelişmiş istem mühendislik tekniklerini ele alıyoruz.
Temel örnekler eğlenceli olsada, bu bölümde daha karmaşık ve ilginç görevler gerçekleştirmemize olanak sağlayan daha gelişmiş istem mühendislik tekniklerini ele alıyoruz.
<ContentFileNames section="techniques" lang="tr"/>

@ -1,5 +1,9 @@
# 提示技术
import ContentFileNames from 'components/ContentFileNames'
时至今日,改进提示显然有助于在不同任务上获得更好的结果。这就是提示工程背后的整个理念。
尽管基础示例很有趣,但在本节中,我们将介绍更高级的提示工程技术,使我们能够完成更复杂和有趣的任务。
尽管基础示例很有趣,但在本节中,我们将介绍更高级的提示工程技术,使我们能够完成更复杂和有趣的任务。
<ContentFileNames section="techniques" lang="zh"/>

Loading…
Cancel
Save