MemPro: Agentic Memory Systems as Evolvable Programs

2026-05-30 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Software Development & Engineering · Depth: Expert, quick

Summary

MemPro introduces a system-level evolution framework designed to overcome limitations in existing agentic memory systems for long-horizon autonomous agents. Current memory construction-retrieval (MCR) pipelines often remain fixed post-deployment, struggling with diverse task-specific failures and evolving memory bank structures. MemPro addresses this by treating the entire MCR pipeline as an evolvable program, maintaining a version tree of implementations. An Evolving Agent iteratively selects promising versions, diagnoses recurring failures, and refines them into improved child versions. Experiments on LongMemEval, LoCoMo, HotpotQA, and NarrativeQA demonstrate that MemPro consistently outperforms static and prompt-level evolving baselines within a few iterations, shows continuous improvement, and achieves a favorable performance-cost trade-off. Code is publicly available.

Key takeaway

For AI Architects designing long-horizon autonomous agents, relying on fixed memory construction-retrieval pipelines will likely lead to performance bottlenecks and misalignment with evolving memory banks. You should consider implementing evolvable memory systems like MemPro, which dynamically adapt the entire pipeline through iterative diagnosis and refinement. This approach promises superior, continuously improving performance and better cost efficiency compared to static or prompt-level evolving baselines.

Key insights

Evolving the entire memory construction-retrieval pipeline as a program, not just the memory bank, significantly enhances long-horizon autonomous agents.

Principles

• Memory systems require dynamic adaptation beyond fixed pipelines.
• Treat the full MCR pipeline as an evolvable program.
• Iterative failure diagnosis guides memory system refinement.

Method

MemPro maintains a version tree of runnable memory-system implementations. An Evolving Agent iteratively selects versions, diagnoses recurring failures, and creates improved child versions through failure-mode-guided edit-debug refinement.

In practice

• Implement system-level evolution for agent memory.
• Use failure-mode-guided refinement for pipeline edits.
• Explore version trees for memory system implementations.

Topics

• Agentic Memory Systems
• Evolvable Programs
• Autonomous Agents
• Memory Construction-Retrieval
• System Evolution Frameworks

Code references

• wanghai673/MemPro

Best for: Research Scientist, AI Engineer, NLP Engineer, AI Scientist, Machine Learning Engineer, AI Architect

Related on AIssential

• AI tools for hearing difficulties — helpful or harmful for language learning? — AI assistive technology presents a dilemma for language learners with hearing difficulties, balancing aid with skill development.
• How Enterprise Marketing Teams Are Actually Using AI Agents in 2026 — Enterprise marketing is shifting from human-assisted AI tools to autonomous AI agents that redesign entire workflows.
• [AINews] Microsoft Build: MAI-Thinking-1 and MAI Family models — Microsoft's MAI models showcase a transparent, full-stack AI strategy emphasizing clean data and local agent execution.
• Are we getting better at explaining things because of our interaction with LLMs? — Crafting clear LLM prompts may enhance human explanatory and communication abilities.
• OpenAI expands Codex with role-specific plugins to build a general-purpose app for non-developers — OpenAI is transforming Codex into a general-purpose, role-specific AI work app, expanding beyond coding for non-developers.
• How to Use Claude Managed Agents? — Anthropic's Claude Managed Agents simplifies AI agent deployment by providing a hosted, stateful, and secure execution environment.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.