A-Mem: Agentic Memory for LLM Agents
Open accessThe DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index; 11 of the 17 sources its claims cite were checked for support.
Abstract
This paper proposes a novel agentic memory system for LLM agents that can dynamically organize memories in an agentic way following Zettelkasten note-taking principles with dynamic indexing and linking.
Overview
CoChat Summary
TL;DR
A-MEM is an agentic memory system for LLM agents that, inspired by the Zettelkasten method, builds structured notes with LLM-generated keywords, tags, and contextual descriptions, then autonomously links and evolves them as new memories arrive. Across six foundation models on LoCoMo and DialSim, it outperforms LoCoMo, ReadAgent, MemoryBank, and MemGPT while using far fewer tokens.
Key findings
- On LoCoMo with GPT-4o-mini, A-MEM reaches 27.02 F1 / 20.09 BLEU-1 on Multi Hop and 45.85 / 36.67 on Temporal, versus MemGPT's 26.65 / 17.72 and 25.52 / 19.44; A-MEM's average ranking is 1.2 versus 2.4 for both LoCoMo and MemGPT.
- On DialSim, A-MEM scores F1 3.45, BLEU-1 3.37, ROUGE-L 3.54, ROUGE-2 3.60, METEOR 2.05, SBERT 19.51, versus LoCoMo (2.55, 3.13, 2.75, 0.90, 1.64, 15.76) and MemGPT (1.18, 1.07, 0.96, 0.42, 0.95, 8.54) — a 35% F1 improvement over LoCoMo and 192% over MemGPT.
- Ablation with GPT-4o-mini: removing both link generation and memory evolution drops Multi Hop F1 to 9.65; removing only memory evolution gives 21.35; full A-MEM gives 27.02.
- Cost/scaling: ~1,200 tokens per memory operation (85–93% reduction vs. ~16,900 for LoCoMo/MemGPT), under $0.0003 per operation, 5.4 s average with GPT-4o-mini and 1.1 s with local Llama 3.2 1B; retrieval time grows only from 0.31 µs to 3.70 µs from 1,000 to 1,000,000 memories.
Methods
- Design: each memory note is `{content, timestamp, keywords, tags, context, embedding, links}`; an LLM generates keywords/tags/context, a text encoder (all-minilm-l6-v2) produces embeddings, top-k nearest neighbors are retrieved by cosine similarity, and an LLM decides link generation and memory evolution (updating neighbors' context, keywords, tags).
- Data: LoCoMo (7,512 QA pairs, dialogues averaging 9K tokens over up to 35 sessions; single-hop, multi-hop, temporal, open-domain, adversarial) and DialSim (multi-party dialogues from Friends, The Big Bang Theory, The Office; 1,300 sessions, ~350,000 tokens, >1,000 questions per session).
- Evaluation: F1 and BLEU-1 as primary metrics, plus ROUGE-L, ROUGE-2, METEOR, and SBERT similarity; baselines LoCoMo, ReadAgent, MemoryBank, MemGPT; six foundation models (GPT-4o-mini, GPT-4o, Qwen2.5-1.5B/3B, Llama 3.2 1B/3B) plus DeepSeek-R1-32B, Claude 3.0 Haiku, and Claude 3.5 Haiku in the appendix; top-k retrieval with k=10 default and category-specific tuning.
Limitations
- Stated by the paper: memory organization quality may still depend on the underlying LLM's capabilities, and different LLMs may generate different contextual descriptions or connections; the implementation is text-only, with multimodal (image/audio) extension left to future work.
- Evident from the text: no error bars or significance tests are reported (the checklist answers "No" to statistical significance, citing API cost); the main-text results are reported for a single dataset family of long-term conversation QA, and the paper's own checklist notes broader societal impacts are not discussed.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
5 Valid · 0 Not valid · 3 Not checked · 55 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“These systems require agent developers to predefine memory storage structures, specify storage points within the workflow, and establish retrieval timing.”
What the sources say
- Partly supportedcites [25] [25] Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G Patil, Ion Stoica, and Joseph E Gonzalez. Memgpt: Towards llms as operating systems. arXiv preprint arXiv:2310.08560, 2023. sourceExperimental — accuracy not yet measured
Section 3.2.1 (p. 6) says the fixed-context baselines fetch documents independently of LLM inference, supporting the idea that retrieval is set by the workflow rather than chosen by the agent. Section 2.3 (p. 3) contrasts this with MemGPT’s self-directed memory edits and retrieval. The paper does not establish that existing memory systems generally require developers to predefine storage structures or specify storage points. (LLM, conf=0.78)
- Could not verifycites [39] [39] Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. Memorybank: Enhancing large language models with long-term memory. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19724–19731, 2024. 12 Contents sourceExperimental — accuracy not yet measured
Sections 2.1 and 2.2 describe MemoryBank’s storage structure and retrieval mechanism, but do not say that existing systems require agent developers to predefine structures, choose workflow storage points, or set retrieval timing. Searches for those developer requirements found no passage addressing them. (LLM, conf=0.73)
- Not matchedcites [28] [28] Aymeric Roucher, Albert Villanova del Moral, Thomas Wolf, Leandro von Werra, and Erik Kaunismäki. 'smolagents': a smol library to build great agentic systems. https://github. com/huggingface/smolagents, 2025.
Its entry has no title and year that we can search by.
- Could not verifycites [21] [21] Zhiwei Liu, Weiran Yao, Jianguo Zhang, Liangwei Yang, Zuxin Liu, Juntao Tan, Prafulla K Choubey, Tian Lan, Jason Wu, Huan Wang, et al. Agentlite: A lightweight library for building and advancing task-oriented llm agent system. arXiv preprint arXiv:2402.15538, 2024. sourceExperimental — accuracy not yet measured
Section 2.1 describes when AgentLite retrieves an action-observation history and stores an observation; the Figure 1 description in Section 6 also describes its memory workflow. Neither passage says that existing memory systems require developers to predefine storage structures, storage points, or retrieval timing. Searches for those requirements found no passage addressing them. (LLM, conf=0.72)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Existing memory systems [25, 39, 28, 21] for LLM agents provide basic memory storage functionality.”
What the sources say
- Supportedcites [25] [25] Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G Patil, Ion Stoica, and Joseph E Gonzalez. Memgpt: Towards llms as operating systems. arXiv preprint arXiv:2310.08560, 2023. source
MemGPT is an LLM-agent memory system that stores information outside the context window. Section 2 (p. 2) describes its external context and memory-management functions, and the Figure 3 description (p. 3) identifies archival and recall storage databases. The claim does not say these systems provide only basic storage. (LLM, conf=0.88)
- Supportedcites [39] [39] Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. Memorybank: Enhancing large language models with long-term memory. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19724–19731, 2024. 12 Contents source
MemoryBank is an LLM memory system with a memory-storage component. Section 2 describes that component as its primary data repository, and §2.1 says it stores conversation records, event summaries, and user information. This supports citing MemoryBank as an example of a system that provides memory storage functionality. (LLM, conf=0.91)
- Not matchedcites [28] [28] Aymeric Roucher, Albert Villanova del Moral, Thomas Wolf, Leandro von Werra, and Erik Kaunismäki. 'smolagents': a smol library to build great agentic systems. https://github. com/huggingface/smolagents, 2025.
Its entry has no title and year that we can search by.
- Supportedcites [21] [21] Zhiwei Liu, Weiran Yao, Jianguo Zhang, Liangwei Yang, Zuxin Liu, Juntao Tan, Prafulla K Choubey, Tian Lan, Jason Wu, Huan Wang, et al. Agentlite: A lightweight library for building and advancing task-oriented llm agent system. arXiv preprint arXiv:2402.15538, 2024. source
Section 2.1, “Individual Agent” (page not available), describes AgentLite’s Memory component as storing an agent’s action-observation chains and saving new actions and observations during execution. This supports citing AgentLite as an example of an LLM-agent system with basic memory storage functionality. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“While previous datasets contain dialogues with around 1K tokens over 4-5 sessions, LoCoMo features much longer conversations averaging 9K tokens spanning up to 35 sessions”
§ 4 Experiment > 4.1 Dataset and Evaluation
What the sources say
- Supportedcites [22] [22] Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. Evaluating very long-term conversational memory of llm agents. arXiv preprint arXiv:2402.17753, 2024. source
The Introduction (p. 2) reports that LoCoMo conversations average 9K tokens and span up to 35 sessions. Its comparison table (p. 2) lists earlier multi-session datasets at about 1,055–1,226 tokens across four or five sessions, supporting the stated contrast. (LLM, conf=0.94)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“T-SNE Visualization of Memory Embeddings Showing More Organized Distribution with A-MEM (blue) Compared to Base Memory (red) Across Different Dialogues.”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“with each memory operation costing less than $0.0003 when using commercial API services”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
25 more claims
Create a free account to see all 63 claims and every source verdict.
- Probably not a substantial claim. 16
- It may state two separate claims as one. 20
- The quote only partly states this claim. 2
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source doesn’t support the claim as written.
In the citing paper
“Relative to the O (1) dictionary lookup used by all prior systems [4], [5], [6], [7], [8], [9], lineage-gated retrieval trades at most 13 . 82 µ s in absolute terms for the safety guarantee of Theorem 1-a tradeoff Experiments 1-2 show buys elimination of an 18 . 8 -25 . 5% structural leak rate at a 13 -14 percentage-point reuse cost.”
From Lineage-Aware Memory Governance: A Derivation-Gated Framework for Privacy-Preserving Column-Level Access Control in Enterprise AI Agents (2026) · cites this paper as [6]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This emerging class of systems extends the capabilities of traditional AI Agents by enabling multiple intelligent entities to collaboratively pursue goals through structured communication [144-146], shared memory [147,148], and dynamic role assignment [22].”
From AI Agents vs. Agentic AI: A Conceptual taxonomy, applications and challenges (2026) · cites this paper as [147]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“from Wang et al. [55], Xu et al. [45] build memory to improve long context performance for specific workflows which suggest that diverse evaluation workflows are needed for better testing of generalizable memory components.”
From MEMTRACK: Evaluating Long-Term Memory and State Tracking in Multi-Platform Dynamic Agent Environments (2025) · cites this paper as [45]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Some other approaches introduce external memory modules ( e.g., summarizers or retrievers) [63, 29, 13, 57], but these are typically trained separately and cannot be optimized end-to-end with the agent's policy. This also introduces additional engineering overhead, as engineers must manage and integrate two separate models.”
From MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents (2025) · cites this paper as [57]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Some others, such as HiAgent[5], A-Mem[23], MemAgent [28] use hierarchical designs to separate working buffers from episodic and long-term stores to relieve context pressure and improve persistence.”
From Learning Hierarchical Procedural Memory for LLM Agents through Bayesian Selection and Contrastive Refinement (2025) · cites this paper as [23]
Experimental — accuracy not yet measured
9 more claims about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ APPENDIX
“Using GPT-4o-mini, A-MEM achieves a ROUGE-L score of 44.27 in Multi-Hop tasks, more than doubling LoComo's 18.09.”
§ APPENDIX
“LOCOMO 9.64 23.92 2.01 18.09 3.40 11.58 26.48 40.20 60.46 69.59 READAGENT 2.47 9.45 0.95 13.12 0.55 5.76 2.99 9.92 6.66 9.79 MEMORYBANK 1.18 5.43 0.52 9.64 0.97 5.77 1.64 6.63 4.55 7.35 MEMGPT 10.58 25.60 4.76 25.22 0.76 9.14 28.44 42.24 36.62 43.75 A-MEM 10.61 25.86 21.39 44.27 3.42 12.09 29.50 45.18 42.62 50.04”
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ APPENDIX
“The Multi-Hop category showcases particularly striking results, where Qwen2.5-15b with A-MEM achieves a ROUGE-L score of 27.23, dramatically surpassing LoComo's 4.68 and ReadAgent's 2.81 - representing a nearly six-fold improvement.”
§ APPENDIX
“Multi Hop Temporal Open Domain Single Hop Adversial RGE-2 RGE-L RGE-2 RGE-L RGE-2 RGE-L RGE-2 RGE-L RGE-2 RGE-L”
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ APPENDIX
“This significant advantage maintains consistency across other metrics, with METEOR scores of 23.43 versus 7.61 and SBERT scores of 70.49 versus 52.30.”
§ APPENDIX
“Multi Hop Temporal Open Domain Single Hop Adversial ME SBERT ME SBERT ME SBERT ME SBERT ME SBERT”
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ 4 Experiment > 4.3 Empricial Results
“our A-MEM demonstrates superior performance in Multi-Hop tasks achieves at least two times better performance”
§ 4 Experiment > 4.1 Dataset and Evaluation
“MEMGPT 26.65 17.72 25.52 19.44 9.15 7.44 41.04 34.34 43.29 42.73 2.4 2.4 16,977 A-MEM 27.02 20.09 45.85 36.67 12.14 12.00 44.65 37.06 50.03 49.47 1.2 1.2 2,520”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 10 20 30 40 50
“44.55 25.43 28.31 32.12 34.32 37.02”
§ 4 Experiment > 4.1 Dataset and Evaluation
“A-MEM 27.02 20.09 45.85 36.67 12.14 12.00 44.65 37.06 50.03 49.47 1.2 1.2 2,520”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 4
- [1][1] Sönke Ahrens. How to Take Smart Notes: One Simple Technique to Boost Writing, Learning and Thinking. Amazon, 2017. Second Edition.
- [5][5] Satanjeev Banerjee and Alon Lavie. Meteor: An automatic metric for mt evaluation with improved correlation with human judgments. In Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, pages 65–72, 2005.
- [18][18] Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33:9459–9474, 2020.
- [19][19] Chin-Yew Lin. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81, 2004.
Cited by
Items 601–700 of 1,030
Tan, Z., Gao, J., & Wang, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.17911 · Cited by 1
Karunanidhi, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.21230 · Cited by 2
“Persistent memory architectures for LLM agents are an active area [12, 16], evaluated primarily on retrieval quality.”
Dual-Layer Agentic Memory with Fast Write Routing and Slow ConsolidationPreprint
Li, W., Nie, D., Lan, R., Lyu, T., Wang, P., Hong, L., Pan, W., Pan, B., Pan, B., & Hu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.22215 · Cited by 1
Wang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.23653 · Cited by 1
EviGraph: Towards Verifiable Evidence Construction for Information-Seeking AgentsPreprint
Chen, J., Mao, Y., & Que, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.24667 · Cited by 1
“Multimodal research agents such as OpenSearch-VL [Chen et al.(2026a)] combine search, browsing, and OCR, while memory systems such as Mem0 [Chhikara et al.(2025)], MemGPT [Packer et al.(2024)], A-MEM [Xu et al.(2025)], and REAL [Lu et al.(2026)] compress long histories for later retrieval.”
SlideBank: A Persistent Hierarchical Evidence Bank for Consistent Whole-Slide ReasoningPreprint
Zhao, B., Huang, G., Zhang, C., Li, A., Tan, Y., Zhou, C., Wang, G., Gao, Z., & Li, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.00342 · Cited by 2
Making Prospective Memory SLM-Shaped: Typed Intention Stores for Small-Model AgentsPreprint
Zhao, J., & Wu, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.01272 · Cited by 1
VeriPhy: Agentic Physical Reasoning for World Model Evaluation and RefinementPreprint
Xu, W., Zhu, Y., Ge, C., Shen, X., Shi, J., Kuen, J., Chen, Y., Tao, M., McComb, C., Gutiérrez, N. G., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.03153 · Cited by 1
“Orthogonally, a fast-growing agent-memory literature asks where an agent’s accumulated experience should live—episodic transcripts, distilled semantic priors, or reusable procedures [6, 30, 36, 41, 48, 57, 79, 81, 89, 90].”
Self-citationGrounding Agent Memory: Environment-Probing Curation for Enterprise AgentsPreprint
Suresh, S., Mak, H., Bhatnagar, S., Methani, C., & Munoz, A. G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.11060 · Cited by 1
Studying Without a Syllabus: Task-Agnostic Environment PreprocessingPreprint
Samuel, V., Ursekar, V., Kalmath, V. S., Shanker, A., Chatrath, V., & Xue, Y. (2026) · arXiv.org · doi:10.48550/arxiv.2609.10824 · Cited by 1
Chen, Y., Zhan, Y., Lu, Z., & Jin, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.14062 · Cited by 1
GraMRAG: Orchestrating Multi-Agent Multi-Step Reasoning via Graph Memory with Reinforcement LearningPreprint
Wang, Z. (2026) · arXiv (Cornell University) · Cited by 1
EchoPath: Execution-Level Replayable Memory for GUI AgentsPreprint
Zhao, Y., Shanmugham, A., Roy, S., & Xu, Y.-X. (2026) · arXiv.org · doi:10.48550/arxiv.2609.16635 · Cited by 2
“…context across hierarchical storage tiers [26], Generative Agents organizes an episodic stream over which reflection and retrieval operate [27], MemoryBank couples storage with Ebbinghaus-inspired forgetting curves [28], and A-MEM maintains dynamically linked, evolving agentic memories [29].”
PACMI: Provenance-Aware Cascading Memory Invalidation for Long-Term LLM AgentsPreprint
Wang, Y., Jiaqi, L. I. U., Zhang, J., Wu, Z., Duan, Y., Zheng, M., & Cai, T. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7480138 · Cited by 1
Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic DesignPreprint
Du, H., Yan, L., Flores, C., & Kadav, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.22086 · Cited by 1
“A-MEM (Xu et al., 2025) moves toward an adaptive memory substrate in which newly added memories are dynamically indexed and linked to existing memories and can trigger updates to their contextual representations.”
Self-Healing Harness for Runtime Oversight of Agent Self-ModificationPreprint
Tayebati, S., Kumar, D., Darabi, N., Krishnan, R., & Trivedi, A. R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.24130 · Cited by 2
RPMem: Learning Long-Term Recurrent Parametric Memory Across Sessions for LLM AgentsPreprint
Zhao, F., Cao, R., Dong, L., Yao, F., Xu, J., Jiang, G., Zhao, Y., Li, Y., & Zhang, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.23466 · Cited by 1
“Adaptive memory organization further links related entries and revises existing records as new information arrives [10].”
Jia, W., Cheng, X., Zhang, Z., Fu, D., Li, L., Chen, W., Wu, Y., & Jin, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32522 · Cited by 1
“Mem0 consolidates facts, A-Mem links contextual notes, and Zep maintains temporal knowledge graphs (Chhikara et al., 2025; Xu et al., 2025; Rasmussen et al., 2025).”
廖文樂, Wang, Z., Zhang, J., Jin, J., Xu, Y., & Dou, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34134 · Cited by 1
Retrieved but Not Delivered: Multimodal Memory Delivery for Long-Term AgentsPreprint
Jiang, Y., Liao, Q., Yin, K., Liu, X., Cuomo, L., & Bacci, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32590 · Cited by 1
“Research has concentrated on S and R (Packer et al., 2023; Park et al., 2023; Chhikara et al., 2025; Xu et al., 2025).”
Just-In-Time Agent Memory with Runtime Agentic ResearchPreprint
Yan, B., Li, C., Qian, H., Lu, S., Li, C., & Liu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34385 · Cited by 1
Simple Agentic Memory for Generalist Robot PoliciesPreprint
Zhang, Y., Zhang, Y., Li, M., Wang, J., Jin, Z., Liu, S., & Zhao, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36595 · Cited by 1
“General agent memories study what to store, consolidate, and retrieve for later reasoning (Packer et al., 2023; Xu et al., 2025; Niu et al., 2026).”
Self-citationWang, X., Wang, S., & Liu, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39740 · Cited by 1
“External memory systems (Packer et al., 2023; Xu et al., 2025a; Chhikara et al., 2025; Li et al., 2025c; Du, 2026) have progressed from preserving evidence beyond the model’s active context to learning how stored information should be retained, retrieved, and transformed for subsequent reasoning.”
MemCodex: Self-Programming Hierarchical Memory for Language AgentsPreprint
Wang, X., & Liu, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39765 · Cited by 1
“Hand-designed baselines are MemGPT (Packer et al., 2023), A-Mem (Xu et al., 2025), Mem0 (Chhikara et al., 2025), MemoryOS (Kang et al., 2025) and HMO (Liu et al., 2026c).”
Causal Memory Policy: Making Memory Utility Identifiable by Intervening on RetrievalPreprint
Behnam, A., & Wang, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.02070 · Cited by 1
Interpreting at Write Time: A Policy Ablation for Multi-Goal Agent MemoryPreprint
Sadowski, A., & Chudziak, J. A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.02897 · Cited by 1
Lu, H., & Gong, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.05303 · Cited by 1
“In-context methods store reflections, memories, or reusable skills and retrieve them at inference [32, 59, 42, 37, 26, 57, 10, 49, 58], so reuse depends on retrieval and on the frozen policy executing it over a long interaction.”
Memory Type Varies: Empowering LLM Agents for Long-Term Memory with Diverse StrategiesPreprint
Wen, Y., Xu, D., Jia, P., Wang, Y., Zhang, Y., Wang, M., Li, J., Zhang, W., Li, X., Liu, Y., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.11573 · Cited by 1
“(1) HyDE [Gao et al., 2023] (2) Mill [Jia et al., 2024] (3) Query2Doc [Wang et al., 2023a] (4) Secom [Pan et al., 2025] (5) A-Mem [Xu et al., 2025c] (6) HippoRAG 2 [Gutiérrez et al., 2025] (7) Raptor [Sarthi et al., 2024] (8) MemoryOS [Kang et al., 2025] (9) LightMem [Fang et al., 2025a].”
InfluentialSelf-citationFrom Existing Methods to New Research Ideas: Two Perspectives on Recursive Self-ImprovementPreprint
Zhou, L.-J., Yu, Y.-K., Liu, Z.-H., Xiang, Q.-J., Hu, T.-C., Dong, Y.-R., Wu, Y.-H., Xiang, Y.-X., Wu, X.-J., Chen, X., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.14625 · Cited by 1
“A-MEM and ReasoningBank illustrate Online improvement by incorporating new records or distilled strategies as tasks are processed [21, 20].”
InfluentialMemCut: Every Secret Leaves a Trail in the Long-Term Memory of LLM AgentsPreprint
Zhang, C., Cao, X., Zheng, Z., & Xiong, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.21405 · Cited by 1
“Beyond the five memories evaluated here, other agent memories (Park et al., 2023; Wang et al., 2023; Xu et al., 2025; Rasmussen et al., 2025; Gutiérrez et al., 2024; Zhu et al., 2026; Park, 2026) vary the same two properties (§6.2).”
InfluentialContinual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM AgentsPreprint
Zhao, H.-Y., Tu, S.-T., Huang, Y., Zhou, H.-C., Yin, Z., Huang, J.-H., Xu, J.-Y., Shi, T.-M., Shi, J.-S., Zhang, N., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.33956 · Cited by 2
“Harness-based learning includes retaining information in long-term memory [12, 47, 81], extracting lessons through reflection [58, 89, 94], and developing reusable skills [64, 70, 73, 95].”
Self-citationGueddes, A., Louhichi, B., & Mahjoub, M. A. (2026) · Cluster Computing · doi:10.1007/s10586-026-06436-8
“Consequently, the field has rapidly shifted from pure retrieval augmentation towards structured, persistent Memory-Augmented Generation (MAG) , where the agent actively manages an organized external memory rather than passively consuming long token sequences [11–13].”
Wang, M., Han, K., Wang, R., Zhang, S., Zhao, Z., & Xu, X. (2026) · International Conference on Big Data and Information Analytics (BigDIA) · doi:10.1109/bigdia71496.2026.11684708
“Retrieval-focused memory improves long-horizon continuity [7], [8]; VAIL adds an authorization layer, neutral on trap-free public tasks and decisive on the targeted modes [17].”
Anthony, C. R. O., & Peiris, S. M. (2026) · Moratuwa Engineering Research Conference (MERCon) · doi:10.1109/mercon71835.2026.11691559
“Some of these systems also rely on LLMs to introduce meaningful links based on similarities making the memory more probabilistic and to have high latency as the LLM must reason to create or update links [8].”
NeurPIU: Neurobiologically Inspired Personalized Intent Understanding in Large Language Models
Liao, Z., Liao, J., & Zhao, X. H. (2026) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3805712.3809593
“Memory-augmented methods equip LLMs with external memory to support long-term interactions [21, 37].”
Gao, Q., Gu, F., & Ji, Y. (2026) · International Conference on Artificial Intelligence, Internet of Things and Cloud Computing Technology (AIoTC) · doi:10.1109/aiotc71250.2026.11688658
“The present agent-memory research argues that memory should not be treated as passive retrieval alone; instead, memory can be actively reconstructed through graph-structured associations between cues, tags, and contents during reasoning [15-17].”
Self-Reflective LLM Agents with Quality-Aware Episodic Memory for Multi-Step Task Completion
Mohanam, B. N. (2026) · International Conference on Cybersecurity, Digital Forensics, and AI Applications (ICCSDFAI) · doi:10.1109/iccsdfai70505.2026.11648054
“A-MEM [7] improved on this with atomic note structures and inter-note links, reporting strong multi-hop F1 on LoCoMo [5].”
Zou, D., Xu, J., Luo, L., & Yu, H. (2026) · International Workshop on Quality of Service · doi:10.1109/iwqos70441.2026.11661050
Cognitively-Inspired Dual Memory for Long-Term Contextual Reasoning in Language Agents
Tyagi, S., Rai, S., Goel, P. K., Garg, D., & Mahur, L. S. (2026) · International Conference on Green Technologies and Sustainable Solutions (GTSS) · doi:10.1109/gtss68743.2026.11679625
An Agentic Approach to Sustainable Environmental Monitoring Systems in Healthcare
Turcu, C. E., & Turcu, C. O. (2026) · IGI Global eBooks · doi:10.4018/979-8-3373-5636-5.ch003
Liu, Y. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-11477-8_11
Vatsal, S., Dubey, H., & Bajaj, A. (2026) · IEEE Access · doi:10.1109/access.2026.3720396
FinPerMA: A Theory-Informed, Event-Grounded Personalized-Memory Benchmark for LLM Agents
Wang, B., Zhou, K., Guo, L., Chen, F., & Zhang, C. (2026) · arXiv · doi:10.48550/arxiv.2608.04095 · Cited by 4
Beyond Context Window: Dynamic Knowledge Graphs for Lifelong Learning Llms
Desai, V., Kalshetty, R., M, M. S., & Kotagi, S. (2026) · IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT) · doi:10.1109/conecct68978.2026.11700592
“More recent work (2024–2025) has explored agent-centric memory architectures such as A-MEM [23] and MemoryOS [25], which organize memory through associative note linking or OS-style memory tiers.”
A survey on AI agent security: Reasoning, Acting, and Self-Evolving
Liu, D., Luo, T., Huang, S., Yang, J., Liang, S., Jia, X., Lau, C. P., Dong, J., Liu, Y., Bai, Y., et al. (2026) · AI Plus · doi:10.59717/ipj.aiplus.2026.100004 · Cited by 3
MARE: Framework to Enhance Long-Term Reasoning with Controlled Memory Augmentation
Porbandarwala, M., Seth, M., Pai, V., Nayak, K., & Deshpande, H. (2026) · International Conference on Innovation & Growth in Next-Gen Intelligent Technology & Engineering (IGNITE) · doi:10.1109/ignite69592.2026.11706478
Agentic Memory: Enabling Situational Resonance in Human-Robot Collaboration
Si, J., Li, W., Li, J., Lv, J., Peng, T., & Bao, J. (2026) · Procedia CIRP · doi:10.1016/j.procir.2026.02.208
AIGuardian: A Parent-Regulated AI Chatbot Interface for Children
Bhullar, S. S., Khan, M., & Rani, K. M. S. (2026) · International Conference on Emerging Techniques in Computational Intelligence (ICETCI) · doi:10.1109/icetci71772.2026.11710429
“Persistent memory for LLM agents is an active area: A-Mem structures agentic memory as linked notes [8], MemOS treats memory as an operating-system abstraction [9], Zep organizes histories in a temporal knowledge graph [10], and Mem0 targets production-grade long-term memory [11].”
From Flat Storage to Structured Graphs: Adaptive Memory for Domain-Specific LLM Reasoning
Erick, M., & Bogacsovics, G. (2026) · IEEE Conference on Information Technology and Data Science (CITDS) · doi:10.1109/citds70571.2026.11706949
“Other related approaches include RAG-based augmentation [3] and agent memory systems such as ExpeL [7], A-MEM [8], and Agent-Fly [9], all of which store experiences as flat, independent entries.”
Abeke, O., Nassr, R. M., & Mahmoud, H. (2026) · International Journal of Data Science and Analytics · doi:10.1007/s41060-026-01290-8
RealMem: Benchmarking LLMs in Real-World Memory-Driven Interaction
Bian, H., Yao, Z., Hu, S., Xu, Z., Zhang, S., Guo, Y., Yang, Z., Han, X., Wang, H., & Chen, R. (2026) · arXiv · doi:10.48550/arxiv.2601.06966
“Each instance contains a user query paired with responses from four systems (MemoryOS, Graph Memory, A-mem, and Mem0), forming 30 four-way comparative evaluation tasks.”
InfluentialAgent-Dice: Disentangling Knowledge Updates via Geometric Consensus for Agent Continual Learning
Wu, Z., Lou, X., Ma, X., Li, Y., Liu, W., Zhang, W., Wang, J., & Zhang, Z. (2026) · arXiv · doi:10.48550/arxiv.2601.03641
“Previous work has primarily approached agent continual learning from two perspectives: agentic reinforcement learning (Zhang et al., 2025a) and agentic memory (Xu et al., 2025a).”
LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues
Wu, D., Ji, Z., Kawatkar, A., Kwan, B., Gu, J.-C., Peng, N., & Chang, K.-W. (2026) · arXiv · doi:10.48550/arxiv.2605.12493
“AgentRunbook-R is an efficient retrieval-augmented generation (RAG) pipeline inspired by agentic memory works such as Xu et al. [2025].”
RecMem: Recurrence-based Memory Consolidation for Efficient and Effective Long-Running LLM Agents
Dai, Z., Deng, S., Guan, S., Tian, Y., Yao, X., Yan, X., & Cheng, J. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.1619
“Most existing memory systems construct long-term memory by incrementally transforming in-coming interactions (or short windows thereof) into retrievable memory units, such as summaries (Kang et al., 2025; Packer et al., 2024; Zhong et al., 2023), atomic facts (Chhikara et al., 2025; Wang and Chen, 2025), or structured…”
InfluentialCognis: Context-Aware Memory for Conversational AI Agents
Daftari, P., Patel, K., Kapale, S., George, J., & Surendira, S. (2026) · arXiv · doi:10.48550/arxiv.2604.19771
“…include: et al., 2025], Zep [Zep AI, 2024], MemGPT [Packer et al., 2023], MemoryBank [Zhong et al., 2023], ReadAgent [Lee et al., 2024], A-Mem [Xu et al., 2025], Cross-benchmark validation on LongMemEval [Wu et al., 2025] confirms these results generalize: Cognis achieves up to 92.4% overall…”
InfluentialMao, W., Tan, H., Liu, S., Liu, H., Xu, Y., Ji, H., & Wang, X. (2026) · arXiv · doi:10.48550/arxiv.2601.06490
“…evaluate the performance of Bi-Mem against multiple leading baselines thoroughly, including: LongContext (LLM backbone without memory), RAG (Lewis et al., 2020), Mem0 (Chhikara et al., 2025), LightMem (Fang et al., 2025), A-MEM (Xu et al., 2025), SeCom (Pan et al., 2025), and CAM (Li et al., 2025).”
InfluentialWhat Deserves Memory: Adaptive Memory Distillation for LLM Agents
Ma, W., Nan, J., & Wu, W. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.1607 · Cited by 89
“N EMORI can be implemented with third-party management systems (Xu et al., 2025; Kang et al., 2025) Generate narrative and cue ( N j , c j ) ← f LLM ( P nar ∥ P j ) Retrieve candidates from D e and decide merge-or-insert 6: Obtain episodic memory M in (merged M ν or new M j ) % — Semantic Knowledge…”
Hippocampus: An Efficient and Scalable Memory Module for Agentic AI
Li, Y., Cao, L., Ahmed, F., Sharma, P., & Li, B. (2026) · arXiv · doi:10.48550/arxiv.2602.13594
“Hybrid systems like A-Mem (Xu et al., 2025) combine both approaches to balance semantic richness and structural precision.”
InfluentialJust-In-Time Reinforcement Learning: Continual Learning in LLM Agents Without Gradient Updates
Li, Y., Lin, Z., Deng, A., Zhang, X., He, Y., Ji, S., Cao, T., & Hooi, B. (2026) · arXiv · doi:10.48550/arxiv.2601.18510
LoCoMo-Plus: Beyond-Factual Cognitive Memory Evaluation Framework for LLM Agents
Li, Y., Guo, W., Zhang, L., Xu, R., Huang, M., Liu, H., Xu, L., Xu, Y., & Liu, J. (2026) · arXiv · doi:10.48550/arxiv.2602.10715
“We evaluate three state-of-the-art conversational memory systems: A-Mem (Xu et al., 2025), which maintains and re-trieves structured long-term memories through adaptive memory construction and retrieval; Mem0 (Chhikara et al., 2025), which provides a production-oriented memory abstraction with scalable long-term stora…”
Zhang, W., Wei, X., Huang, W.-C., Hui, Z., Wang, C., Gong, M., & Yu, P. S. (2026) · arXiv · doi:10.48550/arxiv.2603.25973
“Compared to long-context prompting, Mem0, LoCoMo, and MemoryBank substantially increase ranking scores for GPT-5 and Claude-4 Sonnet, whereas A-Mem is more competitive on rating (e.g., lowest MAE for GPT-5) but does not match the best ranking variants, suggesting a trade-off between decision-time evidence selection an…”
InfluentialAnatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations
Jiang, D., Li, Y., Wei, S., Yang, J., Kishore, A., Zhao, A., Kang, D., Hu, X., Chen, F., Li, Q., et al. (2026) · arXiv · doi:10.48550/arxiv.2602.19320
“Six memory architectures are selected: AMem (Xu et al., 2025c), MemoryOS (Kang et al., 2025a), Nemori (Nan et al., 2025), MAGMA (Jiang et al., 2026a), SimpleMEM (Liu et al., 2026) and MemSkill (Zhang et al., 2026a) as shown in Table 8 of Appendix E.”
InfluentialDMF: A Deterministic Memory Framework for Conversational AI Agents
Stabile, M., & Zimuel, E. (2026) · arXiv · doi:10.48550/arxiv.2606.03463
“A-MEM [7] builds an agentic memory system using Zettelkasten-inspired linking between memory notes, where an LLM assigns metadata and interconnects entries.”
LycheeMemory V2: Efficient Long-Term Memory for LLM Agents via Semantic Segment-Level Consolidation
Li, D., Liu, Z., Wang, J., Huang, J., Li, F., Zhang, L., Jia, B., Hu, B., & Zhang, M. (2026) · arXiv · doi:10.48550/arxiv.2608.12990
Experience Compression Spectrum: Unifying Memory, Skills, and Rules in LLM Agents
Zhang, X., Wang, G., Cui, Y., Qiu, W., Li, Z., Zhu, B., & He, P. (2026) · arXiv
Membox: Weaving Topic Continuity into Long-Range Memory for LLM Agents
Tao, D., Ma, G., Huang, Y., & Jiang, M. (2026) · arXiv · doi:10.48550/arxiv.2601.03785
“Experiments on Lo-CoMo show that membox achieves up to 68% F1 improvement over strong baselines such as Mem0 and A-MEM, while using far fewer context tokens.”
InfluentialSun, X., Wang, Q., Li, H., & Chen, L. (2026) · arXiv · doi:10.48550/arxiv.2608.03137
MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery
Du, S., Yan, X., Shi, J., Cao, Z., Feng, S., Liang, Z., Sun, B., Guo, X., Peng, T., Zhou, Y., et al. (2026) · arXiv · doi:10.48550/arxiv.2606.06473 · Cited by 31
Self-citationSelf-GC: Self-Governing Context for Long-Horizon LLM Agents
Hao, X., Meng, H., Yin, X., Zhu, J., & Cao, C. (2026) · arXiv · doi:10.48550/arxiv.2607.00692
“…as MemGPT, LongMem, Mem0, A-MEM, HippoRAG, REMem, and TiMem further study hierarchical memory, persistent stores, graph-structured memory, and temporal consolidation (Packer et al. 2024; Wang et al. 2023; Chhikara et al. 2025; Xu et al. 2025; Gutiérrez et al. 2024; Shu et al. 2026; Li et al. 2026).”
AriadneMem: Threading the Maze of Lifelong Memory for LLM Agents
Zhu, W., Chen, X., Wang, Z., Wang, J., Dong, X., Huang, M., Cai, R., Sang, H., Wang, H., Qiu, P., et al. (2026) · arXiv · doi:10.48550/arxiv.2603.03290 · Cited by 4
MAPLE-Guard: Memory-Aware Link Enforcement Against Memory-Link Poisoning in Multi-Agent Systems
Xiong, W., Zhou, Y., Wang, J., Gu, S., Tang, B., Li, Z., Xiong, F., Wen, Y., & Wen, M. (2026) · arXiv · doi:10.48550/arxiv.2608.00426 · Cited by 2
“Persistent memory extends retrieval-augmented generation [21] with cross-task experience and hierarchical storage [5, 19, 31, 33, 52, 59].”
MEMAUDIT: An Exact Package-Oracle Evaluation Protocol for Budgeted Long-Term LLM Memory Writing
Bhargava, N., & Barrento, R. S. (2026) · arXiv · doi:10.48550/arxiv.2605.02199
“MemGPT, Reflexion, MemoryBank, Mem0, and A-Mem propose memory architectures or policies for long-term agents [9, 10, 15, 3, 13].”
Lightweight LLM Agent Memory with Small Language Models
Zhang, J., Zhang, C., Chen, S., Huang, Z., Zheng, P., Wang, Z., Guo, P., Mo, F., Bae, S.-H., Zou, J., et al. (2026) · arXiv · doi:10.48550/arxiv.2604.07798
“To maintain cross-turn consistency beyond the context window, many systems augment agents with external memory (Lee et al., 2024; Xu et al., 2025; Hu et al., 2025; Wang et al., 2026).”
InfluentialHeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents
Zhu, J., Li, J., Zhang, C., Liu, J., & Yang, M. (2026) · arXiv · doi:10.48550/arxiv.2604.16839
“Knowledge-organization methods , such as A-Mem (Xu et al., 2025), structure memory into interconnected semantic networks to enable adaptive management.”
InfluentialYu, Y., Yao, L., Xie, Y., Tan, Q., Feng, J., Li, Y., & Wu, L. (2026) · arXiv · doi:10.48550/arxiv.2601.01885
“A-Mem (Xu et al., 2025): We implement A-Mem following the Zettelkasten-inspired design described in the original paper, using the official codebase at https://github.com/WujiangXu/ A-mem-sys/ .”
Influential“Although several memory architectures have been proposed to help agents store and learn from past experiences [49, 40, 51, 59, 65], these efforts primarily focus on using such memories to improve the agent’s accuracy on completing workflows ( e.g., with fewer hallucinations [8] or with higher task…”
EvolMem: A Cognitive-Driven Benchmark for Multi-Session Dialogue Memory
Shen, Y., Pei, D., Guo, Y., Wang, J., Guo, Y., Zhang, Z., Jia, Q., Zhou, J., & Zhai, G. (2026) · arXiv · doi:10.48550/arxiv.2601.03543
“A-MEM introduces an agentic memory architecture in which interaction traces are encoded as LLM-generated “notes” comprising content, times-tamps, keywords, tags, contextual descriptions and embeddings.”
InfluentialDiagnosing Retrieval vs. Utilization Bottlenecks in LLM Agent Memory
Yuan, B., Su, Y., & Yao, K. (2026) · arXiv · doi:10.48550/arxiv.2603.02473
“A growing body of work equips LLM agents with persistent memory [6, 9, 1].”
InfluentialPollertlam, N., & Kornsuwannawit, W. (2026) · arXiv · doi:10.48550/arxiv.2603.04814
MemoryGraft: Persistent Compromise of LLM Agents via Poisoned Experience Retrieval
Srivastava, S. S., & He, H. (2025) · arXiv · doi:10.48550/arxiv.2512.16962 · Cited by 115
“A-Mem[3] organizes memories using Zettelkasten principles.”
Shared Selective Persistent Memory for Agentic LLM Systems
Pedada, S., Dhavala, A., & Patil, N. (2026) · arXiv · doi:10.48550/arxiv.2607.09493 · Cited by 1
“Consider an agentic LLM system that, given a user query q and system prompt s , produces an artifact a through a sequence of tool-use steps T = ( t 1 , t 2 , . . . , t n ) .”
Beyond Similarity: Trustworthy Memory Search for Personal AI Agents
Zhang, J., Chen, K., Ma, J., Hu, Y., He, L., Zhang, Y., Liu, J., Yang, X., Zhang, T., & Jia, R. (2026) · arXiv · doi:10.48550/arxiv.2606.06054 · Cited by 13
“Recent frameworks exemplify this trend: Mem0 [4] extracts and retrieves conversational facts to maintain multi-session coherence; A-Mem [5] structures memories as interconnected notes to model evolving relationships; and MemOS [6] introduces system-level abstractions for managing heterogeneous…”
InfluentialSelf-citationRCR-Router: Efficient Role-Aware Context Routing for Multi-Agent LLM Systems with Structured Memory
Liu, J., Kong, Z., Yang, C., Yang, F., Li, T., Dong, P., Nanjekye, J., Tang, H., Yuan, G., Niu, W., et al. (2025) · arXiv · doi:10.48550/arxiv.2508.04903 · Cited by 37
“A-mem (Xu et al. 2025) proposes a agentic memory system for LLM agents that can dynamically organize memories in an agentic way.”
Accurate and Efficient Long-Term Memory for LLM Agents
Zhao, Z., Guo, X., Lv, L., Wang, M., Li, M., & Li, S. (2026) · arXiv · doi:10.48550/arxiv.2607.16211
“Systems such as MemGPT [Packer et al., 2023], Mem0 [Chhikara et al., 2025], A-Mem [Xu et al., 2025], and MemOS [Li et al., 2025] have demonstrated impressive retention of user preferences, factual history, and contextual nuance across sessions.”
Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents
Zhou, Y., Li, Y., Liu, Z. L., Yavuz, S., & Joty, S. (2026) · arXiv · doi:10.48550/arxiv.2609.27334 · Cited by 1
MemTrace: Probing What Final Accuracy Misses in Long-Term Memory
Long, X., Chen, Z., Zeng, S., Wang, S., Guo, K., & Tang, J. (2026) · arXiv · doi:10.48550/arxiv.2606.17328
The AI Hippocampus: How Far are We From Human Memory?
Jia, Z., Li, J., Kang, Y., Wang, Y., Wu, T., Wang, Q., Wang, X., Zhang, S., Shen, J., Li, Q., et al. (2026) · arXiv · doi:10.48550/arxiv.2601.09113
Are We Ready For An Agent-Native Memory System?
Zhou, W., Zhou, X., Han, S., Xu, H., Li, G., Li, Z., Xiong, F., & Wu, F. (2026) · arXiv · doi:10.48550/arxiv.2606.24775
Deshpande, D., Gangal, V., Mehta, H., Kannappan, A., Qian, R., & Wang, P. (2025) · arXiv · doi:10.48550/arxiv.2510.01353
Chain-of-Memory: Lightweight Memory Construction with Dynamic Evolution for LLM Agents
Xu, X., Xu, B., Xueyun, T., Huang, Z., Chen, R., Yunfan, L., & Shen, H. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.534
Żatuchin, D. (2026) · Research Square (preprint) · doi:10.21203/rs.3.rs-9801639/v1
Locomo-Plus: Beyond-Factual Cognitive Memory Evaluation Framework for LLM Agents
Li, Y., Guo, W., Zhang, L., Xu, R., Huang, M., Liu, H., Xu, L., Xu, Y., & Liu, J. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.1150
Synapse: Empowering LLM Agents with Episodic-Semantic Memory via Spreading Activation
Jiang, H., Chen, J., Pan, Y., Chen, L., You, W., Zhou, Y., Zhang, R., Abate, Y., & Liu, T. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.1108 · Cited by 2
Memory Matters More: Event-Centric Memory as a Logic Map for Agent Searching and Reasoning
Hu, Y., Liu, J., Tan, J., Zhu, Y., & Dou, Z. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.1123
Self-citationEvo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents
Wei, T., Shi, Z., Lin, M., He, B., Liu, Z., Sang, Y., Bei, Y., Ning, X., Zou, J., Li, T.-W., et al. (2026) · arXiv · doi:10.48550/arxiv.2608.15071
“…trajectories offline (Lee et al., 2026; Ni et al., 2026), utilizing media such as reflection (Shinn et al., 2023; Qiu et al., 2026), memory (Zhong et al., 2024; Chhikara et al., 2025; Xu et al., 2025), and skills (Zhang et al., 2026a; Xia et al., 2026; Jiang et al., 2026; Yang et al., 2026).”
MemRefine: LLM-Guided Compression for Long-Term Agent Memory
Kim, M., Baek, J., Jeong, S., & Hwang, S. J. (2026) · arXiv · doi:10.48550/arxiv.2606.13177
“We validate M EM R EFINE on two representative memory frameworks with different representations, namely an A-MEM-style graph memory of linked structured notes (Xu et al., 2025) and the Mem0 pipeline that constructs entries through an ingestion-and-update process (Chhikara et al., 2025), evaluated across multiple LLM j…”
InfluentialAMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications
Zhao, Y., Yuan, B., Huang, J., Yuan, H., Yu, Z., Xu, H., Hu, L., Shankarampeta, A., Huang, Z., Ni, W., et al. (2026) · arXiv · doi:10.48550/arxiv.2602.22769 · Cited by 51
GitSwarm: Decentralized Compounding Inference
Shah, V., Samanta, A., Dahal, P., Plekhanov, M., Wu, C.-J., Yih, S., Munos, R., Fergus, R., Foerster, J., Salakhutdinov, R., et al. (2026) · arXiv · doi:10.48550/arxiv.2610.04862
Graphs Meet AI Agents: Taxonomy, Progress, and Future Opportunities
Bei, Y., Zhang, W., Wang, S., Chen, W., Zhou, S., Chen, H., Li, Y., Bu, J., Pan, S., Yu, Y., et al. (2025) · arXiv · doi:10.48550/arxiv.2506.18019
“For example, A-MEM [89] creates interconnected knowledge networks through dynamic indexing and linking, allowing continuous refinement of contextual representations as new memory is integrated.”
Metadata
- Authors
- Xu, W., Liang, Z., Mei, K., Gao, H., Tan, J., & Zhang, Y.
- Affiliations
- Rutgers, The State University of New Jersey (US)
- Ant Group (China) (CN)
- Salesforce (United States) (US)
- Venue
- Advances in Neural Information Processing Systems 38
- Published
- 2025
- Pages
- 20004-20031
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4407759090
- Cited by
- 51
- References
- 0
- Citations by year
- 2026: 48 · 2025: 3
- FWCI
- 48.02
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- neural information processing systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Information Systems › Information Retrieval and Search Behavior
- Keywords
- agentic memory, LLM agents, knowledge networks, memory organization, dynamic indexing, evolution of memory, knowledge representation, memory management
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.