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 101–200 of 1,030
Memory-R2: Fair Credit Assignment for Long-Horizon Memory-Augmented LLM AgentsPreprint
Yan, S., Bahloul, A., Nie, E., Schwarzmann, S., Trivisonno, R., Tresp, V., & Ma, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.21768 · Cited by 20
ACM: Agentic Context Management for Long Horizon TasksPreprint
Li, X., Ming, R., Chu, M., Shao, S., Jin, R., & Xiong, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.23809 · Cited by 23
“Memory-Augmented approaches (Singh et al., 2025; Packer et al., 2023; Park et al., 2023; Wang et al., 2024a; Xu et al., 2025; Fang et al., 2025) maintain an external memory that is updated continuously—typically through reflection, distillation, or optimization—so that the working context remains…”
TrustMem: Learning Trustworthy Memory Consolidation for LLM Agents with Long-Term Memory
Yang, T., Paul, S., Srinivasan, V., Kulkarni, V., & Chappidi, S. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.2419 · Cited by 20
Omni-SimpleMem: Autoresearch-Guided Discovery of Lifelong Multimodal Agent MemoryPreprint
Liu, J., Ling, Z., Qiu, S., Liu, Y., Han, S., Xia, P., Tu, H., Zheng, Z., Xie, C., Fleming, C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.01007 · Cited by 23
“…(Packer et al., 2023) with OS-inspired memory hierarchies, Generative Agents (Park et al., 2023) with recency-importance-relevance scoring, Simple-Mem (Liu et al., 2026a) with efficient lifelong memory, and A-Mem (Xu et al., 2025a) with LLM-directed reorganization, to multimodal architectures.”
SEAGym: An Evaluation Environment for Self-Evolving LLM AgentsPreprint
Zheng, C., Xue, C., Liang, B., Yang, J., & Zhang, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.17546 · Cited by 20
“Memory-based and skill-based methods build reusable memories, skills, workflow traces, or knowledge bases (Wang et al., 2023, 2024; Xu et al., 2025; Tang et al., 2025).”
LatentMem: Customizing Latent Memory for Multi-Agent Systems
Fu, M., Xue, X., Li, Y., He, Z., Huang, S., Qu, X., Cheng, Y., & Yang, Y. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.457 · Cited by 17
MemMA: Coordinating the Memory Cycle through Multi-Agent Reasoning and In-Situ Self-Evolution
Lin, M., Zhang, Z., Lu, H., Liu, H., Tang, X., He, Q., Zhang, X., & Wang, S. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.2339 · Cited by 19
The evolution of tool use in LLM agents: from single-tool call to multi-tool orchestration
Xu, H., Li, C., Ma, X., Ou, X., Zhang, Z., He, T., Liu, X., Wang, Z., Liang, J., Chu, Z., et al. (2026) · Artificial Intelligence Review · doi:10.1007/s10462-026-11700-9 · Cited by 26
EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement LearningPreprint
Xu, W., Zhao, W., Wang, Z., Li, Y.-J., Jin, C., Jin, M., Mei, K., Wan, K., & Metaxas, D. N. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.22576 · Cited by 18
Self-citationHybrid Self-evolving Structured Memory for GUI AgentsPreprint
Zhu, S., Wu, W., Zhou, K., Wang, S., & Huang, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.10291 · Cited by 17
“To improve organization, graph-based systems (Xu et al., 2025; Zhang et al., 2025a) structure knowledge as connected nodes, enabling agents to perform multi-hop retrieval and reasoning beyond similarity matching (Shen et al., 2025).”
In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents
Tan, Z., Yan, J., Hsu, I.-H., Han, R., Wang, Z., Le, L., Song, Y., Chen, Y., Palangi, H., Lee, G., et al. (2025) · Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2025.acl-long.413 · Cited by 219
Forouzandeh, S., Peng, W., Moradi, P., Yu, X., & Jalili, M. (2025) · arXiv · doi:10.48550/arxiv.2512.18950 · Cited by 16
Liu, B., Li, X., Zhang, J., Wang, J., He, T., Hong, S., Liu, H., Zhang, S., Song, K., Zhu, K., et al. (2025) · arXiv · doi:10.48550/arxiv.2504.01990 · Cited by 16
Uncertainty Quantification in LLM Agents: Foundations, Emerging Challenges, and Opportunities
Oh, C., Park, S., Kim, T. E., Li, J., Li, W., Yeh, S., Du, S., Hassani, H., Bogdan, P., Song, D., et al. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.738 · Cited by 31
WebCoach: Self-Evolving Web Agents with Cross-Session Memory GuidancePreprint
Liu, G., Geng, S., Li, S., Cui, H., Zhang, S., Liu, X., & Liu, T. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.12997 · Cited by 18
“Techniques like latent-space retrieval and Zettelkasten-style linking further extend reasoning across hundreds of thousands of tokens (Wang et al., 2025a; Yu et al., 2025; Xu et al., 2025).”
Agentic Learner with Grow-and-Refine Multimodal Semantic MemoryPreprint
Bo, W., Zhang, S., Sun, Y., Wu, J., Xie, Q., Tan, X., Chen, K., He, W., Li, X., Zhao, N., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.21678 · Cited by 20
“To address the limitations of transient context, a parallel line of research has focused on equipping agents with long-term memory , enabling them to learn from experience and retain knowledge persistently[3, 11, 23, 40, 51].”
Beyond Pipelines: A Survey of the Paradigm Shift toward Model-Native Agentic AIPreprint
Sang, J., Xiao, J., Han, J., Chen, J., Chen, X., Wei, S., Sun, Y., & Wang, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.16720 · Cited by 21
Learning Query-Aware Budget-Tier Routing for Runtime Agent MemoryPreprint
Zhang, H., Yue, H., Feng, T., Long, Q., Bao, J., Jin, B., Zhang, W., Li, X., You, J., Qin, C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.06025 · Cited by 17
Memex(RL): Scaling Long-Horizon LLM Agents via Indexed Experience MemoryPreprint
Wang, Z., Chen, H., Wang, J., & Wei, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.04257 · Cited by 19
“Factual memory functions as a declarative knowledge base that supports consistency (Zhou et al., 2023; Xu et al., 2025; Long et al., 2025), coherence (Liu et al., 2023; Chen et al., 2025b), and logs environment states (Wang et al., 2024a; Jimenez Gutierrez et al., 2024), whereas experiential memory…”
MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI AgentsPreprint
Zeng, Z., Hua, H., Zou, B., Cai, M., Feris, R., & Luo, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.18652 · Cited by 17
“Recent systems further study learned memory control [19, 76], trainable memory operations [62, 83], self-organizing memory frameworks [16, 72], decision-theoretic memory management [49], and efficient compressed or parametric memory representations [4, 36, 41].”
MemConflict: Evaluating Long-Term Memory Systems Under Memory ConflictsPreprint
Tao, Z., Zhao, J., Liu, P., Xi, D., Chen, Y., Xu, W., & Li, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.20926 · Cited by 18
“Beyond representation, memory management also involves operations such as writing, consolidation, abstraction, retrieval, and updating [31, 37].”
Autonomous Research Agents: A Survey of AI Scientists and the Verification GapPreprint
Ding, T., Nannapaneni, A., Liu, B., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.05179 · Cited by 23
Evaluating and regulating agentic AI: A study of benchmarks, metrics, and regulation
Farooq, A., Raza, S., Karim, N., Iqbal, H., Vasilakos, A. V., & Emmanouilidis, C. (2026) · Information Fusion · doi:10.1016/j.inffus.2026.104444 · Cited by 25
HiMA-Ecom: A Long-Horizon Memory Benchmark for Hierarchical Multi-Agent E-commerce AssistantsPreprint
Hu, J., Han, A., Zhan, H., Wei, P., Zhang, Z., Guo, Y., Lu, J., Chen, Z., Li, H., Zhang, Z., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.19846 · Cited by 15
“For memory, RAG-based mechanisms place recalled entries directly into the prompt, where selection [43–45], forgetting [46], and aggregation [47–49] decide what an agent reads.”
A Network Arena for Benchmarking AI Agents on Network TroubleshootingPreprint
Wang, Z., Cornacchia, A., Sacco, A., Galante, F., Canini, M., & Jiang, D. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2512.16381 · Cited by 19
“…[40, 50, 87]; prompting strategies , e.g., how to structure the agent’s reasoning process [34, 43, 66, 70]; tool interfaces , e.g., granularity and abstraction of external APIs [10, 51]; state management , e.g., which historical data to keep in the context at each reasoning step [38, 75].”
VitaBench 2.0: Evaluating Personalized and Proactive Agents in Long-Term User InteractionsPreprint
Chen, Y., Zhang, Y., Cai, Z., Shi, Y., Yao, Z., Cui, C., Zheng, J., 霍雅琪, Su, X., Gu, Q., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.27141 · Cited by 17
“This line of work augments LLMs with external memory mechanisms that support the storage, retrieval, and updating of user-relevant information over time [34, 19, 35].”
Self-citationMemory is Reconstructed, Not Retrieved: Graph Memory for LLM AgentsPreprint
Ji, S., Li, Y., & Hooi, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.06036 · Cited by 19
Omidi, P., Huang, X., Laborieux, A., Nikpour, B., Shi, T., & Eshaghi, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.10824 · Cited by 20
“A-MEM (Xu et al., 2025) advances this through automatically generated and evolving memory notes that form dynamic external knowledge graphs capturing semantic relationships over time.”
InfluentialReinforcement Learning Foundations for Deep Research Systems: A SurveyPreprint
Li, W., Chen, Z., Lin, J., Cao, H., Han, W., Liang, S., Zhang, Z., Dong, K., Li, D., Zhang, C., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.06733 · Cited by 18
KnowMe-Bench: Benchmarking Person Understanding for Lifelong Digital CompanionsPreprint
Wu, T., Chen, Z., Weng, Z., Wang, S., Li, C., Zhang, S., Hu, S., Wu, S., Lan, Q., Wang, H., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.04745 · Cited by 17
“We provide base-lines spanning long-context prompting, retrieval-augmented agents, and external memory-store / agentic-memory systems (Packer et al., 2023; Zhong et al., 2024; Chhikara et al., 2025; Xu et al., 2025).”
InfluentialMulti-Agent Memory from a Computer Architecture Perspective: Visions and Challenges AheadPreprint
Yu, Z., Yu, N., Zhang, H., Ni, W., Yin, M., Yang, J., Zhao, Y., & Zhao, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.10062 · Cited by 17
Belief Memory: Agent Memory Under Partial ObservabilityPreprint
Liao, J., Wang, Q., Zhu, J., Du, B., Yan, R., & Chen, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.05583 · Cited by 17
“Mem0 [Chhikara et al., 2025] dynamically extracts and consolidates salient facts for vector-based retrieval, and A-MEM [Xu et al., 2025] organizes memories as structured notes with indexing and linking.”
Always-OnAgents:A Survey of Persistent Memory, State, and Governance in LLMAgentsPreprint
Ding, T., Nannapaneni, A., Liu, B., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.30306 · Cited by 16
Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System IntelligencePreprint
Feng, Y., Xiang, Z., Yang, C., Ma, Q., Chen, Z., Zhang, Y., Huang, K., Wu, C., Liu, Z., Wang, Y., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.21156 · Cited by 17
Self-citationAtomMem: Learnable Dynamic Agentic Memory with Atomic Memory Operation
Huo, Y., Lu, Y., Zhang, Z., Chen, H., & Lin, Y. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.48550/arxiv.2601.08323 · Cited by 15
“Currently, most memory mechanisms of LLM-based agents rely on static, expert-crafted workflows (Xu et al., 2025; Chhikara et al., 2025; Li et al., 2025).”
Lian, N., Wang, Y., Yao, H., Wang, J., Chen, B., Wang, Y., Zhang, M., & Xia, S. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.533 · Cited by 17
Does Memory Need Graphs? A Unified Framework and Empirical Analysis for Long-Term Dialog Memory
Hu, S., Wei, Y., Ran, J., Han, X., Yao, Z., Wang, H., Chen, R., & Zou, L. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1232 · Cited by 19
Pan, J., Liu, Y., Miao, R., Ding, K., Zheng, Y., Nguyen, Q. V. H., Liew, A. W., & Pan, S. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1407 · Cited by 13
MemIndex: Agentic Event-based Distributed Memory Management for Multi-agent Systems
Saleh, A., Tarkoma, S., Lindgren, A., Donta, P. K., Dustdar, S., Pirttikangas, S., & Lovén, L. (2025) · ACM Transactions on Autonomous and Adaptive Systems · doi:10.1145/3774946 · Cited by 20
“A-Mem[45] ProposeamemorysystemthatenablesLLM .”
A survey of foundation models for IoT: taxonomy and criteria-based analysis
Wei, H., Lee, D. Y., Rohal, S., Hu, Z., Rossi, R. A., Fang, S., & Pan, S. (2025) · CCF Transactions on Pervasive Computing and Interaction · doi:10.1007/s42486-025-00208-w · Cited by 13
From Memory to Skills: Evidence-Grounded Governance for Long-Horizon LLM Agents
Tang, B., Zhang, Y., Zhuang, G., Wei, W., Zheng, G., Xie, L., Tan, Y., Xiong, F., Yang, Q., Chung, E., et al. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.48550/arxiv.2607.16621 · Cited by 13
“Existing memory mechanisms typically store either factual memory , such as raw interactions, user facts, and compressed contexts [3, 16, 34, 44], or experiential memory , such as reflections, reasoning traces, and task summaries [18, 26, 37, 41].”
Self-citationUnderstanding LoRA as Knowledge Memory: An Empirical AnalysisPreprint
Back, S., Lee, D., Kang, N., Lee, T., Hong, S. K., Gwon, Y., & Ahn, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.01097 · Cited by 19
“Beyond retrieval, prior approaches span writable/readable non-parametric stores (Xu et al., 2025; Chhikara et al., 2025) and parametric or architectural mechanisms that internalize updates in weights or structure (Wang et al., 2024; Behrouz et al., 2024; 2025; Dao & Gu, 2024).”
Memory in the LLM Era: Modular Architectures and Strategies in a Unified FrameworkPreprint
Wu, Y., Lin, T., Zhou, Y., Zhang, F., Guo, Q., Zhou, X., Wang, S., Liu, X., Ma, Y., & Fang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.01707 · Cited by 15
Slipstream: Trajectory-Grounded Compaction Validation for Long-Horizon AgentsPreprint
Chen, Z., Pan, R., Dai, Y., & Netravali, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.08580 · Cited by 16
“…accumulated history to an external store and fetch relevant content on demand, ranging from semantic-similarity retrieval over chunked logs [Lewis et al., 2020] to memory-augmented architectures that maintain evolving structured state [Packer et al., 2023, Zhong et al., 2024, Xu et al., 2025].”
Self-Evolving Multi-Agent Systems via Decentralized MemoryPreprint
雅, 好., Long, Y., & Zhao, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.22721 · Cited by 13
“…memory abstractions, including reflective memory formation in Generative Agents [19], scalable personalized memory in Mem0 [3], dynamic memory organization in A-Mem [20], lightweight multi-stage memory management in LightMem [21], and efficient lifelong memory compression in SimpleMem [22].”
Can Agent Memory Systems Track Evolving State?Preprint
Fan, X., Liu, M., Yang, R., Ouyang, S., & Han, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.19652 · Cited by 14
Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents
Liu, Z., Hao, Z., Chen, Y., Wang, H., Hou, J., Ding, R., Yang, Y., Ji, W., Xia, W., & Liu, F. (2026) · arXiv · Cited by 12
“We further compare with additional agent baselines, including A-MEM (Xu et al., 2025), Search-R1 (Jin et al., 2025), and DeepResearcher (Zheng et al., 2025).”
Yao, T., Yang, Y., Yan, Y., Ou, X., Li, M., Wang, C., Li, W., Du, C., Shao, X., Gao, Z., et al. (2025) · AI Agent · doi:10.20517/aiagent.2025.04 · Cited by 15
Agent Memory: Characterization and System Implications of Stateful Long-Horizon Workloads
Omri, Y., Gan, Z., Broveak, Z., Geens, R., He, Z., Pentland, A., Verhelst, M., Weissman, T., & Tambe, T. (2026) · IEEE International Symposium on Workload Characterization · doi:10.1109/iiswc71020.2026.00040 · Cited by 12
Reinforcement Learning for Tool-Integrated Interleaved Thinking towards Cross-Domain GeneralizationPreprint
Chen, Z., Yang, J., Xiao, T., Zhou, R., Zhang, L., Xi, X., Shi, X., Wang, W., & Wang, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.11184 · Cited by 12
“…from different perspectives, such as long-horizon planning (Gao et al., 2025b; Singh et al., 2025; Erdogan et al., 2025), memory management (Xu et al., 2025; Yan et al., 2025; Zhou et al., 2025d), interaction efficiency (Wang et al., 2025; Song et al., 2025), multi-tool selection (Dong et…”
TeleMem: Building Long-Term and Multimodal Memory for Agentic AIPreprint
Chen, C., Guan, M., Lin, X., Li, J., Lin, L., Wang, Q., Chen, X., Luo, J., Sun, C., Zhang, D., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.06037 · Cited by 18
“On the structural front, Mem0 [2], A-Mem [30], and MIRIX [25] propose modular systems that enable self-evolving knowledge consolidation and multi-agent coordination.”
Intent Mismatch Causes LLMs to Get Lost in Multi-Turn ConversationPreprint
Liu, G., Zhu, F., Feng, R., Ma, C., Wang, S., & Meng, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.07338 · Cited by 14
“Systems like Mem0 (Chhikara et al., 2025), A-Mem (Xu et al., 2025), and MemoryBank (Zhong et al., 2024) store user facts to enable long-term contextual awareness.”
The Next Paradigm Is User-Centric Agent, Not Platform-Centric ServicePreprint
Zhang, L., Lv, H., Pan, Q., Wang, K., Huang, Y., Miao, X., Xu, Y., Guo, W., Liu, Y. U., Wang, H., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.15682 · Cited by 12
Combee: Scaling Prompt Learning for Self-Improving Language Model AgentsPreprint
Li, H., He, R., Zhang, Q., Ji, C., Mang, Q., Chen, X., Agrawal, L. A., Liao, W.-L., Yang, E., Cheung, A., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.04247 · Cited by 12
“Dynamic Cheatsheet (Suzgun et al., 2025; Xu et al., 2025; Shinn et al., 2023) demonstrates that compact, evolving textual memory can help language agents adapt at inference time by accumulating reusable guidance from past experience.”
Agent-BRACE: Decoupling Beliefs from Actions in Long-Horizon Tasks via Verbalized State UncertaintyPreprint
Singh, J., Khan, Z., Prasad, A., Chen, J. C.-Y., Nambi, A., Lee, H., Stengel-Eskin, E., & Bansal, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.11436 · Cited by 14
“Several memory-based approaches [Yu et al., 2025, Kang et al., 2025, Zhang et al., 2025, Xu et al., 2025] address the resulting context burden: MEM1 [Zhou et al., 2025] selectively compresses past interactions through summarization.”
RaMem: Contextual Reinstatement for Long-term Agentic MemoryPreprint
Yang, W., Kan, B. K., Li, S., Li, L., Qin, Y., Li, J., Bogdan, P., & Thomason, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.22844 · Cited by 12
“Later work makes these memories more structured by using temporal knowledge graphs, note-like organizations, graph-augmented facts, lightweight memory modules, and multi-view indexes [Rasmussen et al., 2025, Xu et al., 2025, Chhikara et al., 2025, Fang et al., 2025, Liu et al., 2026].”
InfluentialMemAudit: Auditing Long-Term Agent Memory via Hidden User-State RecoveryPreprint
Ma, E., Zhou, Y., Huang, W. E. I.-X., Yang, J., Ma, H., Wang, Z., Li, C., Miao, C., Yu, P. S., & Wang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.24595 · Cited by 14
MemGym: a Long-Horizon Memory Environment for LLM Agents
Xu, W., Wang, Y., Mei, K., Liang, K., Wang, Z., Jin, M., Zhang, H., Zhang, S.-X., Hua, W., Sahu, S., et al. (2026) · arXiv · doi:10.48550/arxiv.2605.20833 · Cited by 11
“We compare seven families: rolling Summary [44], A-Mem [50] (note-evolution), MemoryBank [58], LightMem [11], SimpleMem [27], Naive RAG [25] / BM25 retrieval (M EM G YM -DR only), and a no-memory None control.”
InfluentialSelf-citationAll-Mem: Agentic Lifelong Memory via Dynamic Topology Evolution
Lv, C., Chang, H., Tao, S., Chen, M., Fan, Z., Zhang, Z., Guo, Y., & Zhou, S. (2026) · arXiv · doi:10.48550/arxiv.2603.19595 · Cited by 11
From Recall to Forgetting: Benchmarking Long-Term Memory for Personalized Agents
Uddin, M. N., Shubham, K., Blanco, E., Baral, C., & Wang, G. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.1337 · Cited by 11
A Survey on Autonomy-Induced Security Risks in Large Model-Based Agents
Su, H., Luo, J., Liu, C., Yang, X., Zhang, Y., Dong, Y., & Zhu, J. (2026) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2026.3688650 · Cited by 11
“Recentadvancesinclude schema-sensitive memory (e.g., A-MEM [154]), pruned LLMs for representation control [155], multimodal consistency auditing [156], and immune-style fault healing [149].”
Self-citationMAGNET: Towards Adaptive GUI Agents with Memory-Driven Knowledge Evolution
Sun, L., Zhang, J., Wang, S., & Wei, Z. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1299 · Cited by 13
MemPO: Self-Memory Policy Optimization for Long-Horizon AgentsPreprint
Li, R., Zhang, X., Yu, H., Duan, S., Li, X., Xiang, W., Liao, C., Guo, X., Li, Y., & Suo, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.00680 · Cited by 13
“In recent years, researchers have introduced exter-nal memory systems to address the limitations of LLM context windows (Xu et al., 2025; Chhikara et al., 2025; Zheng et al., 2024; Packer et al., 2024; Zhong et al., 2023).”
InfluentialContextual Agentic Memory is a Memo, Not True MemoryPreprint
Xu, B., Dai, X., & Zhang, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.27707 · Cited by 11
“Let H = { h c : ( a, b ) (cid:55)→ ( ab + c ) mod p | c ∈ Z p } .”
MemPrivacy: Privacy-Preserving Personalized Memory Management for Edge-Cloud AgentsPreprint
Chen, Y., Zhao, J., Tang, B., Wang, H., Zhang, Y., Huang, F., Xiong, F., & Li, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.09530 · Cited by 13
“Moving beyond flat memory stores, A-Mem [45] organizes memory as an evolvable network with dynamic indexing and linking, whereas MemOS [20] reconceptualizes memory as a first-class system resource and provides unified mechanisms for memory representation, organization, and lifecycle governance…”
Self-citationAtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM AgentsPreprint
Wang, P., Hu, Y., Liu, X., Yang, J., Wang, H., & Wen, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.17933 · Cited by 11
“Recent memory-augmented reinforcement learning (RL) systems address this need by storing retrieved skills, retrospectives, and task heuristics across episodes [36, 37, 49, 54].”
Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory
Chhikara, P., Khant, D., Aryan, S., Singh, T., & Yadav, D. (2025) · arXiv · doi:10.48550/arxiv.2504.19413 · Cited by 10
AEL: Evolving Agent Harness in Open-Ended Environments
Xu, W., Han, J., Guo, M., Mei, K., Zhu, X., Zhang, H., & Metaxas, D. N. (2026) · Findings of the Association for Computational Linguistics: EMNLP 2026 · doi:10.18653/2026.findings-emnlp.1316 · Cited by 10
Self-citationMemoria: A Scalable Agentic Memory Framework for Personalized Conversational AI
Sarin, S., Singh, L., Sarmah, B., & Mehta, D. B. (2025) · International Conference on AI-ML Systems (AIMLSystems) · doi:10.1109/aimlsystems67835.2025.11330332 · Cited by 22
“The evaluated methods include: (1) Full Context , where the LLM is prompted without memory augmentation, (2) Memoria , our proposed framework with summarization and KG retrieval, (3) A-Mem (ST) , the original A-Mem setup using SentenceTransformers, and (4) A-Mem (OpenAI) , a modified A-Mem…”
Dynamic Affective Memory Management for Personalized LLM AgentsPreprint
Lu, J., & Li, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.27418 · Cited by 10
A Multi-Memory Segment System for Generating High-Quality Long-Term Memory Content in AgentsPreprint
Zhang, G., Wang, B., Ma, Y., Zhao, D., & Yu, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.15294 · Cited by 11
“MMS has achieved leading performance on multiple main-stream large language models, significantly outperforming existing methods such as NaiveRAG, MemoryBank, and A-MEM.”
InfluentialABBEL: Learning Natural-Language Belief States for Memory-Efficient InteractionPreprint
Lidayan, A., Bjorner, J., Golechha, S., Goyal, K., & Suhr, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2512.20111 · Cited by 12
“Another popular family of methods process long contexts into an external memory store for the agents to query (Packer et al., 2024; Xu et al., 2025; Wang et al., 2023; Zhang et al., 2026); these systems of non-parametric external memory are complimentary to context summarization approaches which…”
AMEM4Rec: Leveraging Cross-User Similarity for Memory Evolution in Agentic LLM RecommendersPreprint
Nguyen, M.-D., Kieu, H.-D., & Le, D. N. H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.08837 · Cited by 12
“Our memory design is inspired by A-Mem [29].”
InfluentialMemCollab: Cross-Model Memory Collaboration via Contrastive Trajectory DistillationPreprint
Chang, Y., Wu, Y., Wu, Q., & Lin, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.23234 · Cited by 16
“Other studies focus on memory organization and operations, such as evolving memory graphs [32] and training-based policies for storage, update, and retrieval [42, 34, 36].”
Back to Basics: Let Conversational Agents Remember with Just Retrieval and GenerationPreprint
Wu, Y., Chen, W., Huang, Z., Chen, J., Liu, Q., Wang, K., Zhou, X., & Liang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.11628 · Cited by 11
PersonaVLM: Long-Term Personalized Multimodal LLMsPreprint
Nie, C., Fu, C., Zhang, Y., Yang, H., & Shan, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.13074 · Cited by 14
“Moreover, while general-purpose memory architectures like A-Mem [43] and Memory OS [22] employ more sophisticated agentic frameworks, their utility in our context is severely constrained.”
Self-citationMemCompiler: Compile, Don't Inject -- State-Conditioned Memory for Embodied AgentsPreprint
Ding, X., Wang, X., Yang, Y., Wu, H., Jiang, S., Zhang, Q., Mi, L., Zhu, H., Li, K., Liu, Y., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.07594 · Cited by 10
“We compare against five memory augmentation baselines spanning graph-structured, retrieval-based, and latent-token approaches: G-Mem [39] and A-Mem [33] organize memory as structured graphs with hierarchical or linked-note representations; Mem0 [13] and LangMem [49] maintain flat memory stores with…”
Remember the Decision, Not the Description: A Rate-Distortion Framework for Agent MemoryPreprint
Zou, M., Guo, Z., Liang, L., Wang, Z., Wang, Q., Wen, Q., King, I., Qu, L., & Xu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.10870 · Cited by 11
WorldMemArena: Evaluating Multimodal Agent Memory Through Action-World InteractionPreprint
Liu, C., Yang, Y., Pu, S. X., Liu, Y., Long, L., Guo, Y., Chen, N., Weng, Z., Kochkina, E., Kaur, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.29341 · Cited by 13
DynamicMem: A Long-Horizon Memory Benchmark in Real-World SettingsPreprint
Xie, W., Zhou, S., Li, Z., Parsa, P., Zhou, S., Ding, X., Arvind, C. S., Wang, G., Braverman, V., Payani, A., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.22877 · Cited by 11
Agent Security Bench (ASB): Formalizing and Benchmarking Attacks and Defenses in LLM-based Agents
Zhang, H., Huang, J., Mei, K., Yao, Y., Wang, Z., Zhan, C., Wang, H., & Zhang, Y. (2024) · arXiv · doi:10.48550/arxiv.2410.02644 · Cited by 9
Self-citationQuery-focused and Memory-aware Reranker for Long Context Processing
Li, Y., Li, J., Yu, M., Chen, Y., Ding, G., Lin, Z., Zhang, W., & Zhou, J. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1607 · Cited by 9
JoyAgent-JDGenie: Technical Report on the GAIAPreprint
Liu, J., Xu, S., Liu, S., Li, Y., Liu, W., Liu, M., Zhou, X., Wang, H., Jia, S., Wang, Z., et al. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.00510 · Cited by 10
“…[104]) pair retrieval-centric planning with iterative decomposition, while memory mechanisms progress from reflective feedback [75] to hierarchical designs like A-Mem [100] that separate working, semantic, and procedural layers—supporting long-horizon continuity and adaptive control.”
AssoMem: Scalable Memory QA with Multi-Signal Associative Retrieval
Zhang, K., Zhang, X., Ahmed, E., Jiang, H., Kumar, C., Sun, K., Lin, Z., Sharma, S., Oraby, S., Colak, A., et al. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.10397 · Cited by 11
“…from existing memory graphs that are built entirely on abstractive concepts rather than raw historical data(Wang et al., 2024; Chhikara et al., 2025; Xu et al., 2025b), this graph supports associative connections from abstractive clues to exact memories, facilitating importance-aware ranking .”
Beyond Parameters: Exploring Virtual Logic Depth for Scaling LawsPreprint
Zhu, R., Zhang, H., Li, K., Shi, T., Duan, Y., Wang, C., Zhou, T., Banerjee, A., & Qin, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2506.18233 · Cited by 9
“Xu et al. (2025) propose A-Mem, an agentic memory framework designed to enhance the long-term reasoning and adaptability of LLM agents.”
TokMem: One-Token Procedural Memory for Large Language Models
Wu, Z., Hao, Y., & Mou, L. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.00444 · Cited by 10
“• RAG : We use Sentence-BERT (Reimers & Gurevych, 2019) to retrieve relevant demonstrations or tool-use examples and prepend them to the query, following memory-augmented generation (Packer et al., 2024; Chhikara et al., 2025; Xu et al., 2025).”
SibylSense: Adaptive Rubric Learning via Memory Tuning and Adversarial ProbingPreprint
Xu, Y., Potje, G., Shandilya, S., Yuan, T., Nunes, L. D. O., Agarwal, R., Asgari, S., Atkinson, A., Kıcıman, E., Lu, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.20751 · Cited by 10
“A-MEM [17] develops a Zettelkasten-inspired agentic memory system that creates structured notes and dynamically indexes, links, and updates them as new information arrives.”
Memanto: Typed Semantic Memory with Information-Theoretic Retrieval for Long-Horizon AgentsPreprint
Abtahi, S. M., Rahnema, R., Patel, H., Patel, N., Fekri, M., & Khani, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.22085 · Cited by 9
“A-MEM [8] adopts a Zettelkasten inspired design in which memories are represented as interconnected notes enriched with contextual metadata.”
MemReranker: Reasoning-Aware Reranking for Agent Memory RetrievalPreprint
Li, C., Zhang, M., Kang, J., Chen, D., Shen, J., Tang, B., Zhou, X., Xiong, F., & Li, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.06132 · Cited by 10
“A-Mem [5] proposes an agentic semantic memory system where memories are structured with contextual tags and dynamically linked.”
Learning, Fast and Slow: Towards LLMs That Adapt ContinuallyPreprint
Tiwari, R., Sareen, K., Agrawal, L. A., Gonzalez, J. E., Zaharia, M., Keutzer, K., Dhillon, I. S., Agarwal, R., & Khatri, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.12484 · Cited by 15
MINTEval: Evaluating Memory under Multi-Target Interference in Long-Horizon Agent SystemsPreprint
Lee, H., Chen, J. C.-Y., Singh, J., Khan, Z., Stengel-Eskin, E., & Bansal, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.18565 · Cited by 9
EMBER: Efficient Memory via Budgeted Evidence Retention for Long-Horizon AgentsPreprint
Li, Y., Banerjee, S., & Che, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.05894 · Cited by 9
“Persistent-memory systems store long-lived interaction records or external memory substrates for later use [Park et al., 2023, Zhong et al., 2023, Packer et al., 2023, Wang et al., 2023, 2024, Chhikara et al., 2025, Anokhin et al., 2024, Xu et al., 2025].”
User as Code: Executable Memory for Personalized AgentsPreprint
Li, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.16707 · Cited by 9
Demystifying Agent Skills: Why They Work-Until They Don'tPreprint
Jiang, Z., Huang, F., Xing, H., Wu, X., Gao, Y., Cao, R., Wang, M., Liu, S., & Li, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.14036 · Cited by 11
“This direction has expanded to structured external memory (Packer et al., 2024; Xu et al., 2025), retrieval-augmented experience reuse (Zhao et al., 2024; Zhou et al., 2025; Zhang et al., 2026b; Wu et al., 2026), procedural-memory management (Belikova et al., 2026), and task-agnostic memory graphs…”
Network Self-Configuration Based on Fine-Tuned Small Language Models
Lira, O. G., Rendón, O. M. C., & da Fonseca, N. L. S. (2026) · IEEE Open Journal of the Communications Society · doi:10.1109/ojcoms.2026.3695460 · Cited by 8
“These systems thus transcend traditional model–tool integrations, embodying self-directed, resilient behavior akin to that of artificial operators rather than passive executors [34].”
Aligning Agentic World Models via Knowledgeable Experience Learning
Ren, B., Yao, Y., Sun, R., Qiao, S., Zhang, N., & Chen, H. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.668 · Cited by 8
ShardMemo: Scope-Before-Routing for Agentic Memory Retrieval
Zhao, Y., Dai, C., Kou, M., Xiu, Y., & Niyato, D. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.48550/arxiv.2601.21545 · Cited by 8
Guo, J., Guo, X., Hu, Y., Long, Z., Sui, X., Zhi, X., Huang, Y., He, H., Zhao, W., Zhao, Y., et al. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1260 · Cited by 11
CoordField: Coordination Field for Agentic UAV Task Allocation in Low-Altitude Urban Scenarios
Zhang, T., Tian, Y., Lin, F., Huang, J., Süli, P. P., Ni, Q., Qin, R., Wang, X., & Wang, F.-Y. (2025) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc60802.2025.11423685 · Cited by 9
“This paradigm is driven by LLMs, and integrates core functionalities such as task decomposition, tool invocation, and multi-agent coordination and scheduling into a cohesive system [8], [9].”
EgoMem: Lifelong Memory Agent for Full-duplex Omnimodal ModelsPreprint
Yao, Y., Yu, N., Li, X., Jiang, X. L., Fang, X., Ma, W., Meng, X., Li, J., Sun, A., & Wang, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.11914 · Cited by 9
“In literature, two primary approaches have been explored to equip textual large language models (LLMs) [39, 6, 42] with long-term memory: extended context windows [49, 50, 55] and memory agents [68, 13, 60, 31].”
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.