Mem$^2$Evolve: Towards Self-Evolving Agents via Co-Evolutionary Capability Expansion and Experience Distillation

2026-04-12

AI Summary

Existing large language model (LLM) agents often treat experience accumulation and asset creation—such as tools or expert sub-agents—as separate processes during self-evolution, leading to limited capability growth and unstable evolution. This work proposes a unified co-evolution framework that integrates both aspects within a single paradigm: experience memory guides the dynamic generation of assets, while newly created assets enable the acquisition of higher-quality experiences, establishing a closed-loop synergy between capability expansion and experience refinement. The framework employs a dual-module architecture comprising experience memory and asset memory, leveraging LLMs, dynamic asset generation, and experience distillation. Evaluated across six task categories and eight benchmarks, it outperforms standard LLMs by 18.53%, experience-only evolving agents by 11.80%, and asset-only evolving agents by 6.46% on average, significantly enhancing both evolutionary efficiency and stability.

Technology Category

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

Application Category

• 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: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

While large language model--powered agents can self-evolve by accumulating experience or by dynamically creating new assets (i.e., tools or expert agents), existing frameworks typically treat these two evolutionary processes in isolation. This separation overlooks their intrinsic interdependence: the former is inherently bounded by a manually predefined static toolset, while the latter generates new assets from scratch without experiential guidance, leading to limited capability growth and unstable evolution. To address this limitation, we introduce a novel paradigm of co-evolutionary Capability Expansion and Experience Distillation. Guided by this paradigm, we propose the \textbf{Mem$^{\textbf{2}}$Evolve}, which integrates two core components: \textbf{Experience Memory} and \textbf{Asset Memory}. Specifically, Mem$^{2}$Evolve leverages accumulated experience to guide the dynamic creation of assets, thereby expanding the agent's capability space while simultaneously acquiring new experience to achieve co-evolution. Extensive experiments across 6 task categories and 8 benchmarks demonstrate that Mem$^{2}$Evolve achieves improvement of 18.53\% over standard LLMs, 11.80\% over agents evolving solely through experience, and 6.46\% over those evolving solely through asset creation, establishing it as a substantially more effective and stable self-evolving agent framework. Code is available at: https://buaa-irip-llm.github.io/Mem2Evolve.

Problem

• self-evolving agents
• co-evolution
• experience distillation
• capability expansion
• asset creation

Innovation

• Co-evolution
• Experience Distillation
• Capability Expansion
• Self-Evolving Agents
• Memory Integration