A Scoping Review of Large Language Models: Architecture and Applications

Abstract

Large Language Models (LLMs) have recently exhibited impressive usability of natural language processing to perform different outcomes. A plethora of research contributions encompassing a wide range of areas, including architectural advances, better training strategies, context length enhancements, fine-tuning, multimodal LLMs, robotics, datasets, benchmarking, efficiency, and more, have been made possible by this accomplishment. It becomes harder to understand the overall progress made in LLM research when methods keep changing at a quick pace and discoveries become commonplace. As methodologies continue to improve and discoveries become more common in LLM research, it has become increasingly difficult to understand the overall progress in this field. A concise but thorough summary of recent advances is essential for the scientific community to have access to, considering the wealth of literature on LLMs. Our careful and autonomous review explores innovative themes at the forefront of LLM research while also diving into pertinent prior concepts.