SelfMem: Self-Optimizing Memory for AI Agents

2026-07-04

AI Summary

Existing AI agents rely on predefined, fixed memory mechanisms that struggle to adapt across diverse tasks and require manual tuning. This work proposes the first self-optimizing memory framework, eliminating handcrafted memory strategies by enabling agents to autonomously explore, evaluate, and iteratively refine their memory usage through feedback signals. Integrating principles of self-improving AI, the framework combines feedback-driven policy exploration, model-guided optimization, and dynamic integration of memory tools. Evaluated on the BEAM benchmark across token budgets of 100K to 1M, the approach achieves official score improvements of 48.7%, 40.8%, and 41.9%, respectively, substantially outperforming existing methods and demonstrating exceptional cross-task generalization and robustness.

Abstract

While current AI agents support increasingly long context windows, tool use, and skill execution for long-horizon tasks, they still require memory systems to effectively leverage historical experience. Existing memory frameworks typically rely on fixed storage, retrieval, and summarization mechanisms, which can be rigid across different tasks and often require manual tuning. To address this limitation, we propose SelfMem, a self-optimizing memory framework. Inspired by prior work on self-improving AI, we follow the principle of "teaching an agent to fish rather than giving it a fish." Instead of forcing the model to follow a predefined memory strategy or format, SelfMem provides an environment with memory tools and feedback signals that allow the agent to explore, evaluate, and refine its own memory strategy. Our results show that SelfMem consistently outperforms retrieval, compression, and agent-memory baselines on BEAM across conversation scales from 100K to 1M tokens. Compared with the strongest baseline, SelfMem improves the official score by 48.7%, 40.8%, and 41.9% at 100K, 500K, and 1M, respectively. Further question-type analysis shows broad robustness across diverse memory demands, and our optimization study shows that model-guided strategy refinement further improves performance.

Problem

• AI agents
• memory systems
• fixed mechanisms
• manual tuning
• long-horizon tasks

Innovation

• self-optimizing memory
• AI agents
• memory strategy
• autonomous refinement
• long-horizon tasks