The Global Cheat Sheet for Prompting Strategies in Generative AI

This article is a combination of images that when combined together forms the ultimate cheat sheet for prompt engineering strategies in Generative AI. This has been created through research and combination of various strategies in use.

Rule 1: Create Context using Personas

To know more about how context has to be created, read this article: https://medium.com/@theaigovernment/creator-persona-consumer-persona-situation-connecting-the-consumer-and-creator-context-b72554cfdce1

Rule 2: Be Clear in Instructions

Follow the tactics given here. Be clear and concise. Make the model guess less, and follow more.

Tactics for Clarity

Rule 3: Use of Reference Texts

Use of content that can be referenced and clear mention of the same enhances the outcomes.

Tactics for Referencing

Rule 4: Break down complex tasks

Make tasks simple. When there are too many steps to a task for the model, always try to break down the task into smaller tasks.

Tactics for Task Simplification

Rule 5: Make Thinking Models

Models need time to think. Models should be instructed to take some time to do reasoning and then come with solution.

Tactics for Thinking Model

Rule 6: Combine different models

Sometimes, one model gives better result for a particular scenario, and another one is better for another. Cross reference the models to get ideal outputs, wherever required.

Tactics for integrating models inside a model

Rule 7: Testing Environment Creation

Always test different prompts, and do so by evaluating the output with ideal answers, in cases where answers are known. So, identify certain situations where you have an ideal answer, and then test the model against the same.

Follow the above rules and refer to them every time you create prompts while developing a model, or for creation of prompts in an already developed model, so as to enhance the same.