Dynamic Affective Memory Management for Personalized LLM Agents

TL;DR

This paper introduces a Bayesian-inspired dynamic memory management system for personalized LLM agents, improving memory relevance and coherence in affective scenarios through entropy minimization.

Contribution

It presents a novel memory update algorithm and a benchmark for emotional expression, enhancing personalization and memory efficiency in AI agents.

Findings

Improved personalization and coherence in LLM agents.

Effective reduction of memory bloat through entropy-based updates.

Superior performance demonstrated on the DABench benchmark.

Abstract

Advances in large language models are making personalized AI agents a new research focus. While current agent systems primarily rely on personalized external memory databases to deliver customized experiences, they face challenges such as memory redundancy, memory staleness, and poor memory-context integration, largely due to the lack of effective memory updates during interaction. To tackle these issues, we propose a new memory management system designed for affective scenarios. Our approach employs a Bayesian-inspired memory update algorithm with the concept of memory entropy, enabling the agent to autonomously maintain a dynamically updated memory vector database by minimizing global entropy to provide more personalized services. To better evaluate the system's effectiveness in this context, we propose DABench, a benchmark focusing on emotional expression and emotional change toward objects. Experimental results demonstrate that, our system achieves superior performance in personalization, logical coherence, and accuracy. Ablation studies further validate the effectiveness of the Bayesian-inspired update mechanism in alleviating memory bloat. Our work offers new insights into the design of long-term memory systems.