How to get better and more precise answers from ChatGPT.

Did you know that you can now receive much better answers from generative AI with just a few simple sentences?

As you may have noticed, language models like ChatGPT and Bard do not always perform as well as you would like.

Luckily for us, three researchers from Mohamed bin Zayed University of AI (UAE) defined 26 principles that enhance the LLM ability to focus on the crucial elements of the input context, leading to the generation of quality responses.

The 26 prompts

The 26 prompts that will help you acquire better and more precise replies are listed below.

According to their unique nature, the researchers grouped them into five categories:

(1) Prompt Structure and Clarity, e.g., integrate the intended audience in the prompt, such as the audience is an expert in the field;

(2) Specificity and Information, e.g., Add to your prompt the following phrase: “Ensure that your answer is unbiased and does not rely on stereotypes.”;

(3) User Interaction and Engagement, e.g., Allow the model to elicit precise details and requirements from you by asking you questions until he has enough information to provide the needed output “From now on, I would like you to ask me questions to …”.

(4) Content and Language Style, e.g., No need to be polite with LLMs; there is no need to add phrases like “please”, “if you don’t mind”, “thank you”, “I would like to”, etc., and get straight to the point;

(5) Complex Tasks and Coding Prompts, e.g., Break down complex tasks into a sequence of simpler prompts in an interactive conversation.

Prompts that lead to the best answers

If you are using a Large Scale LLM, such as ChatGPT, then the principles that yield the best improvement with respect to "normal" prompts are the following:

14) Allow the model to elicit precise details and requirements from you by asking you questions until he has enough information to provide the needed output (for example, “From now on, I would like you to ask me questions to…”).

26) To write any text, such as an essay or paragraph, that is intended to be similar to a provided sample, include the following instructions: “Please use the same language based on the provided paragraph.”

2) Integrate the intended audience in the prompt, e.g., “The audience is an expert in the field.”

Prompts that result in the most correct answers

Instead, when dealing with mathematical or logical questions, the best prompts to employ to receive a correct and precise answer are the following:

3) Break down complex tasks into a sequence of simpler prompts in an interactive conversation.

17) Use delimiters [such as commas (,), semicolon (;), quotes ( “, ‘ ), braces ({}), pipes (|), or slashes ( / \ ) to separate text]

19) Combine Chain-of-thought (CoT) prompting with few-Shot prompts.

Example

The authors present a few examples of how language models improve their answers when receiving a better prompt.

In the snippet below, the authors applied Principle Number 7, “Implement example-driven prompting (use few-shot prompting).”

Conclusion

Thanks to these masterfully crafted prompts, we can say goodbye to the annoying situations in which ChatGPT just does not seem to be collaborating.

Whether you are looking for the correct lines of code or whether you are asking for a summary of a text, use these prompts to receive a better answer. Or, as the paper says, “boosted by 50%”.