Document Type : Original Research Paper

Authors

• Zohre Moteshakker Arani
• Mahdi Naghibi

Faculty of Electrical & Computer Engineering, Malek Ashtar University of Technology, Iran

Abstract

Background and Objectives: Modern LLMs struggle with retaining long-term knowledge, personalized context, and complex reasoning over extended texts. A surge of memory-augmented LLM architectures addresses this by integrating external memory mechanisms. Broadly, these methods include retrieval-augmented generation (RAG) (augmenting LLMs with external documents or knowledge bases), explicit memory modules (learned read/write stores), long-context transformers (architectural extensions for longer inputs), episodic/personal memory (storing user or session history), and neuro-symbolic memory (knowledge-graph or symbolic integration).
Methods: In this survey, we present a technical overview of how short-term, long-term, and other memory mechanisms are implemented in and around LLMs (including multimodal agents), and how these innovations enable more capable AI systems. We survey key journal and peer-reviewed works (2021–2025) in each category, emphasizing advances in reasoning and assistant-style tasks. We compare prominent LLM systems that incorporate memory, discuss applications in agents and continuous dialogues, explore the use of memory in different applications, and highlight challenges and future research directions in memory-augmented LLMs.
Results: We contribute a comparative mapping between memory types and representative application domains—such as conversational assistants, knowledge-intensive reasoning, scientific research support, and personalized systems—to address the central question of which memory strategy best aligns with which task requirements.
Conclusion: Although considerable advances have been achieved, important challenges remain. The literature suggests that increasingly sophisticated memory mechanisms will play a central role in the development of next-generation LLMs and AI agents, necessitating future research directions toward more robust, adaptive, and scalable memory-augmented language model systems.

Keywords

• Large Language Models
• Memory Augmented LLMs
• Retrieval Augmented Generation
• Hybrid Memory Architectures

Main Subjects

• Artificial Intelligence

Open Access

This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit: http://creativecommons.org/licenses/by/4.0/

 

Publisher’s Note

JECEI Publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

 

Publisher

Shahid Rajaee Teacher Training University