Decision-Making in Agentic Frameworks for Large Language Model Applications

Abstract

Recent advances in agentic frameworks, powered by large language models (LLMs), have led to significant developments in decision-making systems
that integrate reasoning and acting. This paper provides a comparative analysis of various decision-making models within agentic frameworks, focusing on reasoning strategies, memory augmentation, and situational awareness. Specifically, I analyze models such as ReAct, memory augmented systems like JARVIS-1, and cognitive language architectures like CoALA. I demonstrate that incorporating memory and situational awareness significantly enhances agentic capabilities in complex environments. My findings contribute to a deeper understanding of the comparative strengths and limitations of these reasoning approaches, providing valuable insights for future autonomous agent design.