Prompt Engineering Guide: Techniques, Papers, Prompts

About the Prompt Engineering Guide

Prompt Engineering Guide is a GitHub repository maintained by dair-ai that gathers guides, technique write-ups, model-specific notes, and papers about prompting large language models. It's organized into sections on introduction basics, techniques, applications, a prompt hub of task examples, model notes, and risks, and the same material is also published as a website at promptingguide.ai.

Why Prompt Engineering Matters

Getting useful output from a large language model rarely comes down to typing a plain request: the same task can produce different results depending on whether you supply examples, ask the model to reason step by step, or restructure the prompt entirely. The README frames prompt engineering as work that helps people understand what LLMs can and can't do, used by researchers pushing model performance on hard reasoning tasks and by developers building dependable interfaces between LLMs and other tools. That's a moving target. Because the techniques, papers, and model-specific quirks behind that work are scattered across blogs, papers, and vendor docs, dair-ai built this guide to collect them in one evolving place.

What the Guide Covers

• ✓A Techniques section with a long list of individual pages, from zero-shot and few-shot prompting through chain-of-thought, self-consistency, generate knowledge prompting, prompt chaining, tree of thoughts, RAG, ART, automatic prompt engineer, active-prompt, directional stimulus prompting, program-aided language models, ReAct, multimodal CoT, and graph prompting.
• ✓An Applications section with write-ups covering function calling, generating data, generating synthetic datasets for RAG, tackling generated-dataset diversity, generating code, and a graduate job classification case study.
• ✓A Prompt Hub of example prompts organized by task, including classification, coding, creativity, evaluation, information extraction, image generation, mathematics, question answering, reasoning, text summarization, truthfulness, and adversarial prompting.
• ✓A Models section with dedicated notes on ChatGPT, Code Llama, Flan, Gemini, GPT-4, LLaMA, Mistral 7B, Mixtral, OLMo, and Phi-2, plus a broader model collection page.
• ✓A Risks and Misuses section on adversarial prompting, factuality, and biases.
• ✓A Papers page grouped into overviews, approaches, applications, and collections, plus separate Tools, Notebooks, Datasets, and Additional Readings pages.
• ✓A recorded 1-hour lecture on prompting techniques, applications, and tools, with a companion Jupyter notebook and slide deck.
• ✓A web version at promptingguide.ai that the README points to for the most current guides.

Popularity and Reach

Per the README, cumulative learners passed the 3-million mark by January 2024, the project climbed to the top spot on Hacker News on February 21, 2023, and the content is now available in 13 languages, with more translations welcomed. Outlets including the Wall Street Journal, Forbes, and Markettechpost have covered it, and the team runs a Discord server, a Twitter account, a YouTube channel, and a newsletter for people tracking the space.

Practical Applications of Prompt Engineering

• •Wiring function calling into an LLM application, using the guide's Applications write-up on function calling as a starting point.
• •Building a RAG pipeline and needing synthetic data for it, covered by the guide's pages on generating synthetic datasets for RAG.
• •Comparing how the same prompt behaves across GPT-4, LLaMA, Gemini, Mistral 7B, and other models via the Models section before committing to one for a project.
• •Pulling a starting prompt from the Prompt Hub for a task like classification, coding, or text summarization instead of writing one from scratch.
• •Auditing a production prompt for adversarial inputs, factuality problems, or bias using the Risks and Misuses section.
• •Onboarding a team on prompting fundamentals with the 1-hour recorded lecture and its notebook instead of pointing them at a stack of links.

Other Prompt Engineering Resources

Common Questions

Who should try it — and who should skip

Reach for Prompt Engineering Guide if you're a developer or researcher who wants a single reference for prompting techniques, model-specific notes, and risk considerations, or a team lead who needs a lecture and notebook to onboard people quickly. Skip it, or treat it as a supplement, if you want a fully guided curriculum with feedback along the way; that experience sits behind the paid DAIR.AI Academy courses rather than in the free repo.

Related repositories