Prompt Engineering: A Comparative Study of Prompting Techniques

Urmila Thakare

Abstract: A Large Language Models (LLMs) are used for problem-solving, reasoning, and for content generation. Performance of these models are depending on the way of instructions are given to models. This research is containing the comparative analysis of the three types of prompts techniques used for AI model, that?s include zero-shot prompt, Chain-of-thought prompt, and role-based prompt technique. The dataset used for this research contains 50 questions that?s include mathematical, logical and basic programming problems. The dataset of 50 questions was checked using three different techniques with ChatGPT. That responses of AI model are used for the check of accuracy of all questions. The results of that experiment shown all the prompt types are used for this research give 100% accuracy for the dataset used. But in this we notice responses of all questions are depends on the how ask question which prompt technique used for this. In this not study only about accuracy of answers that given by AI models it's also study quality of answers. Check quality of answers using different prompting technique are show quality of answers is different.

Keywords: Prompt engineering, Large Language Models, Zero-shot Prompting, Chain-of-thought Prompting, Role-based Prompting