MemArchitect: A Policy Driven Memory Governance Layer

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kumar, Lingavasan Suresh, Ba, Yang, Pan, Rong
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908900243013632
author Kumar, Lingavasan Suresh
Ba, Yang
Pan, Rong
author_facet Kumar, Lingavasan Suresh
Ba, Yang
Pan, Rong
contents Persistent Large Language Model (LLM) agents expose a critical governance gap in memory management. Standard Retrieval-Augmented Generation (RAG) frameworks treat memory as passive storage, lacking mechanisms to resolve contradictions, enforce privacy, or prevent outdated information ("zombie memories") from contaminating the context window. We introduce MemArchitect, a governance layer that decouples memory lifecycle management from model weights. MemArchitect enforces explicit, rule-based policies, including memory decay, conflict resolution, and privacy controls. We demonstrate that governed memory consistently outperforms unmanaged memory in agentic settings, highlighting the necessity of structured memory governance for reliable and safe autonomous systems.
format Preprint
id arxiv_https___arxiv_org_abs_2603_18330
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MemArchitect: A Policy Driven Memory Governance Layer
Kumar, Lingavasan Suresh
Ba, Yang
Pan, Rong
Artificial Intelligence
Human-Computer Interaction
Machine Learning
Multiagent Systems
Persistent Large Language Model (LLM) agents expose a critical governance gap in memory management. Standard Retrieval-Augmented Generation (RAG) frameworks treat memory as passive storage, lacking mechanisms to resolve contradictions, enforce privacy, or prevent outdated information ("zombie memories") from contaminating the context window. We introduce MemArchitect, a governance layer that decouples memory lifecycle management from model weights. MemArchitect enforces explicit, rule-based policies, including memory decay, conflict resolution, and privacy controls. We demonstrate that governed memory consistently outperforms unmanaged memory in agentic settings, highlighting the necessity of structured memory governance for reliable and safe autonomous systems.
title MemArchitect: A Policy Driven Memory Governance Layer
topic Artificial Intelligence
Human-Computer Interaction
Machine Learning
Multiagent Systems
url https://arxiv.org/abs/2603.18330