Memory Papers
Survey

From Human Memory to AI Memory: A Survey on Memory Mechanisms in the Era of LLMs

AuthorsYaxiong Wu, Sheng Liang, Chen Zhang, Yichao Wang

arXiv 20252025
arXivPDF

TL;DR

From Human Memory to AI Memory uses a 3D-8Q memory taxonomy to unify personal and system memory mechanisms across object, form, and time.

SharePost on XLinkedIn

Read our summary here, or open the publisher PDF on the next tab.

THE PROBLEM

Memory in LLM agents lacks a unified view across object, form, and time

Existing work mainly analyzes memory only from the temporal dimension, focusing on short-term and long-term memory distinctions.

From Human Memory to AI Memory argues this ignores object and form, leaving LLM-driven AI systems with fragmented personal and system memory designs.

HOW IT WORKS

Three-dimensional, eight-quadrant memory taxonomy

From Human Memory to AI Memory introduces Personal Memory, System Memory, and the Three-Dimensional Eight-Quadrant Memory Taxonomy as core organizing components for AI memory.

An analogy is a computer with RAM, disk, and user profiles, where each quadrant specifies who the memory is about, how it is stored, and for how long.

This KEY_MECHANISM lets From Human Memory to AI Memory describe memory behaviors that a plain context window cannot, such as long-term episodic personalization versus procedural system skills.

DIAGRAM

Mapping human memory types to AI memory dimensions

This diagram shows how From Human Memory to AI Memory aligns human sensory, working, explicit, and implicit memory with AI personal and system memory across form and time.

DIAGRAM

Survey organization across personal and system memory

This diagram shows how From Human Memory to AI Memory structures surveyed work into personal versus system memory and non parametric versus parametric forms.

PROCESS

How From Human Memory to AI Memory Handles Memory Mechanism Analysis

  1. 01

    Human Memory

    From Human Memory to AI Memory first analyzes Human Memory, detailing short-term, long-term, and Memory Mechanisms like encoding, storage, retrieval, and forgetting.

  2. 02

    Memory of LLM driven AI Systems

    From Human Memory to AI Memory then introduces Memory of LLM driven AI Systems, defining object, form, and time as core dimensions.

  3. 03

    3D 8Q Memory Taxonomy

    From Human Memory to AI Memory proposes the 3D 8Q Memory Taxonomy, mapping personal and system memory into eight quadrants across parametric and non parametric forms.

  4. 04

    Personal Memory and System Memory

    From Human Memory to AI Memory finally surveys Personal Memory and System Memory work, connecting each quadrant to concrete systems like MemoryBank and HippoRAG.

KEY CONTRIBUTIONS

Key Contributions

  • 01

    Systematically define LLM driven AI systems memory

    From Human Memory to AI Memory formalizes memory in LLM-driven AI systems by relating Human Memory categories to Personal Memory and System Memory across object, form, and time.

  • 02

    Three dimensional eight quadrant memory taxonomy

    From Human Memory to AI Memory proposes the Three-Dimensional Eight-Quadrant Memory Taxonomy, organizing memory into eight roles like Episodic Memory and Procedural Memory.

  • 03

    Survey of personal and system memory research

    From Human Memory to AI Memory reviews Personal Memory and System Memory work, covering construction, management, retrieval, usage, and benchmarks such as MADial Bench and LOCOMO.

RESULTS

By the Numbers

Quadrants defined

8 quadrants

3 dimensions object form time

Human memory types

4 types

sensory working explicit implicit

Personal memory stages

4 stages

construction management retrieval usage

Survey sections

2 sections

personal memory and system memory

From Human Memory to AI Memory does not report benchmarks but instead defines 8 quadrants across 3 dimensions and maps 4 major human memory types, proving that SYS_NAME unifies diverse memory mechanisms conceptually.

BENCHMARK

By the Numbers

From Human Memory to AI Memory does not report benchmarks but instead defines 8 quadrants across 3 dimensions and maps 4 major human memory types, proving that SYS_NAME unifies diverse memory mechanisms conceptually.

BENCHMARK

Memory dimensions in the 3D 8Q taxonomy

Relative share of conceptual dimensions used in From Human Memory to AI Memory.

KEY INSIGHT

The Counterintuitive Finding

From Human Memory to AI Memory argues that classifying memory only by time, into short-term and long-term, is insufficient for LLM systems.

This is surprising because many LLM memory papers explicitly frame contributions purely as short-term versus long-term context handling, ignoring object and form dimensions.

WHY IT MATTERS

What this unlocks for the field

From Human Memory to AI Memory unlocks a shared vocabulary to design memory modules that distinguish personal versus system, parametric versus non parametric, and short-term versus long-term.

Builders can now place systems like MemoryBank, HippoRAG, and KV Cache into precise quadrants, making it easier to combine, compare, and extend memory mechanisms systematically.

~14 min read← Back to papers

Related papers

SurveyAgent Memory

Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations

Dongming Jiang, Yi Li et al.

arXiv 2026 · 2026

Anatomy of Agentic Memory organizes agentic memory into four structures using components like Lightweight Semantic Memory, Entity-Centric and Personalized Memory, Episodic and Reflective Memory, and Structured and Hierarchical Memory. Anatomy of Agentic Memory then reports comparative results such as Nemori’s 0.781 semantic judge score on LoCoMo versus SimpleMem’s 0.298, and latency differences like 1.129s for Nemori versus 32.372s for MemoryOS.

arXiv:2602.19320Read explainer
SurveyBenchmarkAgent MemoryLong-Term MemoryMemory Architecture

A Survey on the Security of Long-Term Memory in LLM Agents: Toward Mnemonic Sovereignty

Zehao Lin, Chunyu Li, Kai Chen

· 2026

Mnemonic Sovereignty analyzes long term Write, Store, Retrieve, Execute, Share, and Forget Rollback phases against integrity, confidentiality, availability, and governance objectives for agent memory. Mnemonic Sovereignty’s lifecycle matrix shows most of the ~70 works cluster on write and retrieve integrity, leaving store, availability, and governance primitives like write gate validation and post deletion verification almost entirely unexplored.

arXiv:2604.16548Read explainer
SurveyRAGAgent Memory

Memory for Autonomous LLM Agents:Mechanisms, Evaluation, and Emerging Frontiers

Pengfei Du

· 2026

Memory for Autonomous LLM Agents decomposes agent memory into a POMDP-grounded write–manage–read loop, a three-dimensional taxonomy, and five mechanism families spanning context compression, retrieval stores, reflection, hierarchical virtual context, and policy-learned management. Memory for Autonomous LLM Agents synthesizes results like Voyager’s 15.3× tech-tree speedup and MemoryArena’s 80%→45% drop to show that memory architecture often matters more than backbone choice.

arXiv:2603.07670Read explainer

Questions about this paper?

Paper: From Human Memory to AI Memory: A Survey on Memory Mechanisms in the Era of LLMs

Answers use this explainer on Memory Papers.

Checking…