MemInsight: Autonomous Memory Augmentation for LLM Agents

2025-03-27

AI Summary

To address context performance degradation in LLM agents caused by long-term memory inflation and insufficient semantic structuring, this paper proposes MemInsight, an autonomous memory-enhancement framework. MemInsight introduces the first human-annotation-free semantic memory distillation mechanism, integrating self-supervised memory compression, graph-augmented semantic retrieval, and an LLM-driven memory reflection and reorganization module. It enables dynamic index construction and lightweight cross-task generalization without reliance on external vector databases. Evaluated on LLM-REDIAL, MemInsight improves recommendation persuasiveness by 14%; on LoCoMo, it achieves a 34% higher retrieval recall than RAG baselines. Furthermore, it significantly enhances response accuracy and contextual consistency across three diverse tasks: conversational recommendation, question answering, and event summarization.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language model (LLM) agents have evolved to intelligently process information, make decisions, and interact with users or tools. A key capability is the integration of long-term memory capabilities, enabling these agents to draw upon historical interactions and knowledge. However, the growing memory size and need for semantic structuring pose significant challenges. In this work, we propose an autonomous memory augmentation approach, MemInsight, to enhance semantic data representation and retrieval mechanisms. By leveraging autonomous augmentation to historical interactions, LLM agents are shown to deliver more accurate and contextualized responses. We empirically validate the efficacy of our proposed approach in three task scenarios; conversational recommendation, question answering and event summarization. On the LLM-REDIAL dataset, MemInsight boosts persuasiveness of recommendations by up to 14%. Moreover, it outperforms a RAG baseline by 34% in recall for LoCoMo retrieval. Our empirical results show the potential of MemInsight to enhance the contextual performance of LLM agents across multiple tasks.

Problem

• Enhancing semantic data representation for LLM agents
• Improving retrieval mechanisms in long-term memory
• Boosting accuracy and contextualization in agent responses

Innovation

• Autonomous memory augmentation for LLM agents
• Enhanced semantic data representation and retrieval
• Empirical validation in multiple task scenarios