An Empirical Study of Structured Prompt Engineering in Large Language Models

Abstract

Abstract: Large Language Models (LLMs) have significantly transformed artificial intelligence by enabling advanced reasoning, natural language understanding, and content generation. While architectural advancements and large-scale datasets contribute to their effectiveness, prompt engineering has emerged as a critical factor influencing output quality and reliabilityThis study presents a structured empirical evaluation of five prompting strategies: unstructured, role-based, chain-of-thought (CoT), instructional, and constraint-based prompting. The findings indicate that structured prompting techniques improve accuracy by up to 35% in reasoning-intensive tasks and significantly reduce hallucinations. Key Contributions: Development of a structured evaluation framework for prompt strategies. Comparative analysis across multiple task domains. Quantitative assessment of reasoning and safety improvements.

Ähnliche Arbeiten