UMEM: Unified Memory Extraction and Management Framework for Generalizable Memory

TL;DR

UMEM introduces a unified framework for memory extraction and management in LLM-based agents, improving generalization and performance across benchmarks by jointly optimizing memory processes and evaluating utility across semantic clusters.

Contribution

The paper presents UMEM, a novel self-evolving memory framework that jointly optimizes extraction and management, incorporating semantic neighborhood modeling for better generalization.

Findings

Achieves up to 10.67% improvement in multi-turn tasks.

Maintains monotonic growth during continuous evolution.

Outperforms state-of-the-art baselines on five benchmarks.

Abstract

Self-evolving memory serves as the trainable parameters for Large Language Models (LLMs)-based agents, where extraction (distilling insights from experience) and management (updating the memory bank) must be tightly coordinated. Existing methods predominately optimize memory management while treating memory extraction as a static process, resulting in poor generalization, where agents accumulate instance-specific noise rather than robust memories. To address this, we propose Unified Memory Extraction and Management (UMEM), a self-evolving agent framework that jointly optimizes a Large Language Model to simultaneous extract and manage memories. To mitigate overfitting to specific instances, we introduce Semantic Neighborhood Modeling and optimize the model with a neighborhood-level marginal utility reward via GRPO. This approach ensures memory generalizability by evaluating memory utility across clusters of semantically related queries. Extensive experiments across five benchmarks demonstrate that UMEM significantly outperforms highly competitive baselines, achieving up to a 10.67% improvement in multi-turn interactive tasks. Futhermore, UMEM maintains a monotonic growth curve during continuous evolution. Codes and models will be publicly released.