A-MEM: Agentic Memory for LLM Agents
2025/02/17 by Wujiang Xu, Zujie Liang, Xu, Wujiang +9 · 3 voices · 124 citations
Engineering · Computer Science · #cs.CL #cs.HC
paper · pdf · doi:10.48550/arxiv.2502.12110
Abstract
While large language model (LLM) agents can effectively use external tools for complex real-world tasks, they require memory systems to leverage historical experiences. Current memory systems enable basic storage and retrieval but lack sophisticated memory organization, despite recent attempts to incorporate graph databases. Moreover, these systems' fixed operations and structures limit their adaptability across diverse tasks. To address this limitation, this paper proposes a novel agentic memory system for LLM agents that can dynamically organize memories in an agentic way. Following the basic principles of the Zettelkasten method, we designed our memory system to create interconnected knowledge networks through dynamic indexing and linking. When a new memory is added, we generate a comprehensive note containing multiple structured attributes, including contextual descriptions, keywords, and tags. The system then analyzes historical memories to identify relevant connections, establishing links where meaningful similarities exist. Additionally, this process enables memory evolution - as new memories are integrated, they can trigger updates to the contextual representations and attributes of existing historical memories, allowing the memory network to continuously refine its understanding. Our approach combines the structured organization principles of Zettelkasten with the flexibility of agent-driven decision making, allowing for more adaptive and context-aware memory management. Empirical experiments on six foundation models show superior improvement against existing SOTA baselines. The source code for evaluating performance is available at https://github.com/WujiangXu/A-mem, while the source code of the agentic memory system is available at https://github.com/WujiangXu/A-mem-sys.
Cited by
- Keep It InMind: Benchmarking the Implicit-Association Blind Spot in Agent Memory
- Memex(RL): Scaling Long-Horizon LLM Agents via Indexed Experience Memory
- MemChain: Learning Interpretable Memory Traces for Memory-Augmented LLM Agents
- ACM: Agentic Context Management for Long Horizon Tasks
- ClawRec: A Claw-Native Recommender System
- Co-Evolving Graph and Text Memory for Training-Free Multi-Hop Question Answering
- LazyMem: Retrieve Broadly, Construct Selectively for Efficient Long-Term Agent Memory
- Measuring and Improving Behavioral Consistency in Large Language Models through Fact-Heuristic-Emotion State Enforcement
- RSMeM: Knowledge-Enhanced Memory Evolution for Remote Sensing Agents with Systematic Evaluation
- SF-AMS: Strategic Forgetting for Structured Memory in LLM Agent
- Beyond Heuristics: A Decision-Theoretic Framework for Agent Memory Management
- ABBEL: Learning Natural-Language Belief States for Memory-Efficient Interaction
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents
- MemR3: Memory Retrieval via Reflective Reasoning for LLM Agents
- Explainable and Fine-Grained Safeguarding of LLM Multi-Agent Systems via Bi-Level Graph Anomaly Detection
- A Network Arena for Benchmarking AI Agents on Network Troubleshooting
- MemoryGraft: Persistent Compromise of LLM Agents via Poisoned Experience Retrieval
- CogMem: A Cognitive Memory Architecture for Sustained Multi-Turn Reasoning in Large Language Models
- AgentIAD: Agentic Industrial Anomaly Detection via Adaptive Memory Augmentation
- Beyond Training: Enabling Self-Evolution of Agents with MOBIMEM
- Hindsight is 20/20: Building Agent Memory that Retains, Recalls, and Reflects
- Memoria: A Scalable Agentic Memory Framework for Personalized Conversational AI
- Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution
- Experience-Evolving Multi-Turn Tool-Use Agent with Hybrid Episodic-Procedural Memory
- PersonaMem-v2: Towards Personalized Intelligence via Learning Implicit User Personas and Agentic Memory
- MemLoRA: Distilling Expert Adapters for On-Device Memory Systems
- Personalizing Agent Privacy Decisions via Logical Entailment
- MemVerse: Multimodal Memory for Lifelong Learning Agents
- Network Self-Configuration based on Fine-Tuned Small Language Models
- Real-Time Procedural Learning From Experience for AI Agents
- Agentic Learner with Grow-and-Refine Multimodal Semantic Memory
- Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory
- Improving Language Agents through BREW: Bootstrapping expeRientially-learned Environmental knoWledge
- General Agentic Memory Via Deep Research
- A Simple Yet Strong Baseline for Long-Term Conversational Memory of LLM Agents
- MirrorMind: Empowering OmniScientist with the Expert Perspectives and Collective Knowledge of Human Scientists
- Learning to Debug: LLM-Organized Knowledge Trees for Solving RTL Assertion Failures
- LLM-MemCluster: Empowering Large Language Models with Dynamic Memory for Text Clustering
- O-Mem: Omni Memory System for Personalized, Long Horizon, Self-Evolving Agents
- LiCoMemory: Lightweight and Cognitive Agentic Memory for Efficient Long-Term Reasoning
- Smarter Together: Creating Agentic Communities of Practice through Shared Experiential Learning
- BudgetMem: Learning Selective Memory Policies for Cost-Efficient Long-Context Processing in Language Models
- Cache Mechanism for Agent RAG Systems
- MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning
- Tree Training: Accelerating Agentic LLMs Training via Shared Prefix Reuse
- EvoMem: Improving Multi-Agent Planning with Dual-Evolving Memory
- Dynamic Affective Memory Management for Personalized LLM Agents
- Context Engineering 2.0: The Context of Context Engineering
- MemTX: Transactional Belief Commit for Stateful Agent Memory
- Fewer Clarifications, Better Code: Benchmarking Cross-Session Personalized Ambiguity Adaptation in Coding Assistants
- WikiLoop: Jointly Learning to Build and Navigate Agent-Native Wikis with Downstream Feedback
- Metis: Memory Foundation Model
- Functional Cache Grafting: Robust and Rapid Code-Policy Synthesis for Embodied Agents
- DMF: A Deterministic Memory Framework for Conversational AI Agents
- Structured Belief State and the First Precision-Aware Benchmark for LLM Memory Retrieval
- Can Current Agents Close the Discovery-to-Application Gap? A Case Study in Minecraft
- Understanding LoRA as Knowledge Memory: An Empirical Analysis
- RGMem: Renormalization Group-inspired Memory Evolution for Language Agents
- AgentFold: Long-Horizon Web Agents with Proactive Context Management
- WebLeaper: Empowering Efficiency and Efficacy in WebAgent via Enabling Info-Rich Seeking
- Evaluating Long-Term Memory for Long-Context Question Answering
- LLM-empowered knowledge graph construction: A survey
- Learning from Supervision with Semantic and Episodic Memory: A Reflective Approach to Agent Adaptation
- LightMem: Lightweight and Efficient Memory-Augmented Generation
- MemoryBench: A Benchmark for Memory and Continual Learning in LLM Systems
- MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents
- Can Tool-Integrated Reinforcement Learning Generalize Across Diverse Domains?
- How2: How to learn from procedural How-to questions
- A Survey on Agentic Multimodal Large Language Models
- PISA: A Pragmatic Psych-Inspired Unified Memory System for Enhanced AI Agency
- AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval
- SkillOS: Learning Skill Curation for Self-Evolving Agents
- Autonomous Agents for Scientific Discovery: Orchestrating Scientists, Language, Code, and Physics
- Preference-Aware Memory Update for Long-Term LLM Agents
- Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management
- Code Agent can be an End-to-end System Hacker: Benchmarking Real-world Threats of Computer-use Agent
- The Cognitive Bandwidth Bottleneck: Shifting Long-Horizon Agent from Planning with Actions to Planning with Schemas
- LEGOMem: Modular Procedural Memory for Multi-agent LLM Systems for Workflow Automation
- Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models
- JoyAgent-JDGenie: Technical Report on the GAIA
- TokMem: Tokenized Procedural Memory for Large Language Models
- Mem-α: Learning Memory Construction via Reinforcement Learning
- ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory
- MemGen: Weaving Generative Latent Memory for Self-Evolving Agents
- Agentic Services Computing
- Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents
- Estimating the Empowerment of Language Model Agents
- EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning
- SGMem: Sentence Graph Memory for Long-Term Conversational Agents
- MemTxn: A Transaction Boundary for Source-Supported Updates and Complete-State Recovery in Agent Memory
- MIND: Lightweight and Effective Memory Injection Defense for LLM Agents via Intent-Aware Information Bottleneck
- ConMem: Contribution-Aware Memory for Long-Horizon Manufacturing Inspection Logs
- ChronoMem: Version Control and Semantic Rollback for Large Language Model Agent Memory
- Bridging Inference-Time Scaling and Episodic Memory with Action-Centric Graphs
- SkillSmith: Learning to Compose Parametric Skills and Textual Knowledge
- AutoMem: Automated Learning of Memory as a Cognitive Skill
- Escaping the Self-Confirmation Trap: An Execute-Distill-Verify Paradigm for Agentic Experience Learning
- How LoRA Remembers? A Parametric Memory Law for LLM Finetuning
- LLM-based Agents Suffer from Hallucinations: A Survey of Taxonomy, Methods, and Directions
- EpiCache: Episodic KV Cache Management for Long Conversational Question Answering
- SignalLLM: A General-Purpose LLM Agent Framework for Automated Signal Processing
- From Language to Action: A Review of Large Language Models as Autonomous Agents and Tool Users
- A Scenario-Driven Cognitive Approach to Next-Generation AI Memory
- ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization
- EgoMem: Lifelong Memory Agent for Full-duplex Omnimodal Models
- Text2Mem: A Unified Memory Operation Language for Memory Operating System
- AgentArch: A Comprehensive Benchmark to Evaluate Agent Architectures in Enterprise
- Pre-Storage Reasoning for Episodic Memory: Shifting Inference Burden to Memory for Personalized Dialogue
- Meta-Policy Reflexion: Reusable Reflective Memory and Rule Admissibility for Resource-Efficient LLM Agent
- Memento: Fine-tuning LLM Agents without Fine-tuning LLMs
- Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning
- Memory-Augmented Transformers: A Systematic Review from Neuroscience Principles to Enhanced Model Architectures
- PRELUDE: A Benchmark Designed to Require Global Comprehension and Reasoning over Long Contexts
- Seeing, Listening, Remembering, and Reasoning: A Multimodal Agent with Long-Term Memory
- Intrinsic Memory Agents: Heterogeneous Multi-Agent LLM Systems through Structured Contextual Memory
- Memp: Exploring Agent Procedural Memory
- Training-Free Multimodal Large Language Model Orchestration
- RCR-Router: Efficient Role-Aware Context Routing for Multi-Agent LLM Systems with Structured Memory
- Nemori: Self-Organizing Agent Memory Inspired by Cognitive Science
- A Survey of LLM-based Deep Search Agents: Paradigm, Optimization, Evaluation, and Challenges
- L3M+P: Lifelong Planning with Large Language Models
- DeepSieve: Information Sieving via LLM-as-a-Knowledge-Router
- MemTool: Optimizing Short-Term Memory Management for Dynamic Tool Calling in LLM Agent Multi-Turn Conversations
- Graph-Augmented Large Language Model Agents: Current Progress and Future Prospects
Discussions
Related