MemEvolve: Meta-Evolution of Agent Memory Systems

2025-12-21

AI Summary

Existing LLM-based agents rely on static, hand-crafted memory architectures that lack meta-adaptability to task contexts, severely limiting their evolutionary capability. Method: We propose the first meta-evolutionary framework for memory systems, enabling joint co-evolution of memory architecture and experiential knowledge. Our approach defines a modular, comparable memory design space and employs evolutionary algorithms to jointly optimize encoding, storage, retrieval, and management components. We release EvolveLab—a unified open-source experimental platform integrating 12 diverse memory systems. Contribution/Results: Evaluated across four agent benchmarks, our evolved memory architectures achieve up to 17.06% absolute performance gain. Critically, the learned architectures demonstrate strong cross-task and cross-model generalization: they seamlessly transfer to distinct agent frameworks (e.g., SmolAgent, Flash-Searcher) and various LLM backbones. This work marks the first realization of agents capable of *dynamically optimizing how to learn*, establishing a foundational step toward self-improving intelligent agents.

Technology Category

• Machine Learning: Evolutionary Learning
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

Self-evolving memory systems are unprecedentedly reshaping the evolutionary paradigm of large language model (LLM)-based agents. Prior work has predominantly relied on manually engineered memory architectures to store trajectories, distill experience, and synthesize reusable tools, enabling agents to evolve on the fly within environment interactions. However, this paradigm is fundamentally constrained by the staticity of the memory system itself: while memory facilitates agent-level evolving, the underlying memory architecture cannot be meta-adapted to diverse task contexts. To address this gap, we propose MemEvolve, a meta-evolutionary framework that jointly evolves agents' experiential knowledge and their memory architecture, allowing agent systems not only to accumulate experience but also to progressively refine how they learn from it. To ground MemEvolve in prior research and foster openness in future self-evolving systems, we introduce EvolveLab, a unified self-evolving memory codebase that distills twelve representative memory systems into a modular design space (encode, store, retrieve, manage), providing both a standardized implementation substrate and a fair experimental arena. Extensive evaluations on four challenging agentic benchmarks demonstrate that MemEvolve achieves (I) substantial performance gains, improving frameworks such as SmolAgent and Flash-Searcher by up to $17.06%$; and (II) strong cross-task and cross-LLM generalization, designing memory architectures that transfer effectively across diverse benchmarks and backbone models.

Problem

• Evolves memory architecture alongside agent knowledge
• Addresses static memory limitations in agent evolution
• Enables cross-task and cross-model memory adaptation

Innovation

• Meta-evolutionary framework jointly evolves memory architecture and knowledge
• Unified codebase distills memory systems into modular design space
• Achieves performance gains and strong cross-task generalization