Performance Evaluation and Comparative Ranking of LLM Variants in Entity Relationship Prediction

This paper presents a comprehensive evaluation and ranking of five Large Language Models - Mistral, Vicuna, Phi, Llama 2, and Zephyr - across five temperature variants. The focus is on their ability to predict the likelihood of a relationship between pairs of entities within a given context. Each LLM variant is assessed based on stability over multiple epochs, speed, dispersion of results, and the number of incorrect responses. The study aims to provide insights into the performance of these LLMs, thereby facilitating their effective utilization in generating meaningful tabular data. The findings are expected to contribute significantly to the field of artificial intelligence and data analysis. The paper concludes with a comparative ranking of the 25 LLM variants, providing a clear perspective on their relative performances.