MemEvolve: Meta-Evolution of Agent Memory Systems

Guibin Zhang, Haotian Ren, Chong Zhan, Junhao Wang, He Zhu, Wangchunshu Zhou, Shuicheng Yan
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:158215-158234, 2026.

Abstract

Self-evolving memory systems are rapidly 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 work and promote openness in future self-evolving systems, we introduce EvolveLab, a unified memory codebase that distills twelve representative memory systems into a modular design space (encode, store, retrieve, manage), providing a standardized implementation substrate and a fair experimental arena. Extensive evaluations on four challenging agentic benchmarks show that MemEvolve delivers (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, yielding memory architectures that transfer effectively across diverse benchmarks and backbones.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-zhang26fa, title = {{M}em{E}volve: Meta-Evolution of Agent Memory Systems}, author = {Zhang, Guibin and Ren, Haotian and Zhan, Chong and Wang, Junhao and Zhu, He and Zhou, Wangchunshu and Yan, Shuicheng}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {158215--158234}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/zhang26fa/zhang26fa.pdf}, url = {https://proceedings.mlr.press/v306/zhang26fa.html}, abstract = {Self-evolving memory systems are rapidly 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 work and promote openness in future self-evolving systems, we introduce EvolveLab, a unified memory codebase that distills twelve representative memory systems into a modular design space (encode, store, retrieve, manage), providing a standardized implementation substrate and a fair experimental arena. Extensive evaluations on four challenging agentic benchmarks show that MemEvolve delivers (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, yielding memory architectures that transfer effectively across diverse benchmarks and backbones.} }
Endnote
%0 Conference Paper %T MemEvolve: Meta-Evolution of Agent Memory Systems %A Guibin Zhang %A Haotian Ren %A Chong Zhan %A Junhao Wang %A He Zhu %A Wangchunshu Zhou %A Shuicheng Yan %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-zhang26fa %I PMLR %P 158215--158234 %U https://proceedings.mlr.press/v306/zhang26fa.html %V 306 %X Self-evolving memory systems are rapidly 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 work and promote openness in future self-evolving systems, we introduce EvolveLab, a unified memory codebase that distills twelve representative memory systems into a modular design space (encode, store, retrieve, manage), providing a standardized implementation substrate and a fair experimental arena. Extensive evaluations on four challenging agentic benchmarks show that MemEvolve delivers (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, yielding memory architectures that transfer effectively across diverse benchmarks and backbones.
APA
Zhang, G., Ren, H., Zhan, C., Wang, J., Zhu, H., Zhou, W. & Yan, S.. (2026). MemEvolve: Meta-Evolution of Agent Memory Systems. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:158215-158234 Available from https://proceedings.mlr.press/v306/zhang26fa.html.

Related Material