ChatGPT is very popular, and the most comprehensive prompt engineering guide is on the GitHub hot list, with 4.7k stars!

Prompt project can be said to be a "required course" for playing with AI models such as ChatGPT, DALL·E 2, etc.

But how to write this "prompt" specifically, there is some mystery in it...

It’s no wonder that a new profession was born - reminder engineer, with an annual salary of US$250,000-330,000.

Not long ago, Kevin Liu, a Chinese undergraduate from Stanford University, passed prompt The injection method allows all prompts searched by Microsoft ChatGPT to be leaked. Since then, there has been a wave of teasing ChatGPT.

Today, a new project was released on GitHub, which organized the tips Engineering guides, papers, lectures and resources are the most comprehensive prompt package in history.

Project address: https://github.com/dair-ai/Prompt-Engineering-Guide

With a growth of nearly 1k stars a day, this "Prompt Engineering Guide" has been on both GitHub Trending and Hacker News hot lists.

Guide content

The output of a language model needs to be achieved through input hints, but the quality of the results depends on how much information you provide it.

As we introduce more and more examples and applications of prompt engineering, you will notice that a prompt is usually composed of several different elements:

• Instructions: Specific tasks or instructions you want the model to perform
• Context: Supplementary external or contextual information that can guide the model to respond better
• Input data: Input or problem you want to solve
• Output indication:The type or format of the output

Exactly what type or format the output should be How to play with the prompt project? This guide on GitHub will be of great help.

This guide includes readings such as lectures, prompt introductions, papers, tools and libraries, data sets, blogs, and tutorials, with a total of 6 parts.

50 pages of PPT, one-hour super detailed lecture

in " The "Lecture" section includes an hour-long video, code examples, and a 50-page PPT to accompany the lecture.

Among them, the video contains 4 parts, namely, an introduction to the project, providing engineering technology, tools and applications, and future directions.

Both the video and PPT introduce the definition of prompts in detail: prompts refer to the instructions and context passed to the language model in order to achieve a specific task.

Prompt engineering is the process of creating a set of prompts or questions to guide users to obtain their desired results.

Why is the prompt project so important?

Because it is very important for research and discovery, it can be used to help test the various limits of large language models, and can also develop various innovative applications based on large language models.

## Why is prompt engineering so important?

Because it is very important for research and discovery, it can be used to help test the various limits of large language models, and can also develop various innovative models based on large language models. application.

Prompt Guide

Prompt engineering guide developed by the team, mainly composed of 5 parts:

Introduction
• Basic tips

Text summary
• Question answers
Dialogue
• Code generation
• Inference

Advanced tips
• few-shot tips

• Chain of Thought (CoT) Tips

• zero-shot CoT

• Self-consistency

• Generate knowledge prompts

• Automatic Prompt Engineer (APE)

• Adversarial prompts

• Ignore previous commands

• Prompt leak

• Jailbreak

• Others

• Procedurally Assisted Language Model

• ReAct

• Multimodal CoT Tips

• Picture tips

##Collection of papers

The team will be updating the latest papers on prompt engineering every day, and summaries of these papers will be incorporated into the guide above every week.

For example, LeCun’s new work "Augmented Language Models: a Survey" in February this year was included.

Specifically, the author divided the paper into 4 major categories:

Survey/Overview

• Methods/Techniques
• Application
• Supplement

##Tool Kit

data set

Blogs, Guides, Tutorials and Others

This part is mainly a summary of experiences from the big guys.

Author introduction

Elvis Saravia, who leads this project, is DAIR Co-founder of .AI.

He received his master's and doctoral degrees in information systems and applications from Tsinghua University in Taiwan.

Prior to this, he served as a technical product marketing manager at Meta AI for 2 years and was a member of the program committee for the NeurIPS Symposium and NAACL 2019.

Judging from the experience on Linkedin, this friend seems to have been responsible for Galactica, which LeCun has always been thinking about.

The above is the detailed content of ChatGPT is very popular, and the most comprehensive prompt engineering guide is on the GitHub hot list, with 4.7k stars!. For more information, please follow other related articles on the PHP Chinese website!