Hallucination is one of the prominent challenges faced by LLMs in generating responses, hence, reducing the reliability and trustworthiness of LLMs across various domains. Effectiveness of these models highly depend on how they are prompted. While prior work explores individual prompting techniques, this research shows a comparative analysis of multiple free tier LLMs using seven prompting techniques-Zero-shot, One-shot, Few-shot, Tree-of-Thought, Chain-of-Thought, Thread-of-Thread and Graph-of-Thought in a unified framework. The authors evaluate the performance of these combinations on three metrices- accuracy, hallucination rate and response time, offering practical insights and empirical analysis for prompting strategies in real-world applications.

Large Language Models Hallucination Prompting Retrieval-Augmented Generation (RAG)