MRMS: A Multi-Resolution Memory Substrate for Long-Lived AI Agents
2026-07-06 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors propose a special memory system to help AI agents remember important past experiences over time, instead of just relying on short chat history. Their design organizes memory in two ways: by how information is stored (like records or graphs) and by how long it's kept (short-term to long-term). This system helps the AI decide what memories to use, check for conflicts, and update its knowledge when things change. They built a simple prototype to test these ideas during ongoing interactions where remembering and adjusting to new evidence is important.
AI agentsmemory architecturevector representationsgraph relationsshort-term memorylong-term memorymemory consolidationevidence attributioncontextual memorymemory revision
Authors
Jizhizi Li, Amy Shi-Nash
Abstract
Long-lived AI agents require continuity across interactions, but continuity cannot be obtained by simply extending the prompt window. An agent must preserve useful prior experience, retrieve it selectively, distinguish personal context from external evidence, and revise memory when the underlying situation changes. We propose an architectural memory substrate organized along two orthogonal axes: a representational axis spanning structured records, vector representations, and graph relations; and a temporal axis spanning short-term traces, medium-term abstractions, and long-term semantic commitments. Its key design constraint is synchronized structured-vector-graph memory: structured records govern eligibility, vector representations support recall, and graph relations adjudicate support, contradiction, and supersession before gated context projection. Its central claim is that reliable personalization is a memory design problem: useful memory is structured, selectively exposed, continuously consolidated, and epistemically labeled rather than stored as undifferentiated conversation history. Beyond the framework, we instantiate MRMS as a lightweight prototype implementing structured records, vector retrieval, temporal policies, and graph-based revision. The prototype exercises the core substrate mechanisms through pre-generation memory selection, revision, boundary enforcement, and evidence attribution under controlled long-lived interaction scenarios with explicit evidence requirements.