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 401–500 of 1,030
Enhancing Conversational Agents via Task-Oriented Adversarial Memory AdaptationPreprint
Deng, Y., Fu, Y., Xu, D., Wang, Y., Ni, W., Gao, J., Li, X., Liu, C., Han, X., Zhao, G., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.21797 · Cited by 4
“• A-MEM [40] is a memory system that constructs structured notes • LightMem [6] combines the three-level human memory model”
Learning to Share: Selective Memory for Efficient Parallel Agentic SystemsPreprint
Fioresi, J., Kulkarni, P. P., Vayani, A., Wang, S., & Shah, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.05965 · Cited by 6
“…Tan et al., 2025; Zhang et al., 2025c; Chatter-jee & Agarwal, 2025; Fang et al., 2025; Nan et al., 2025; Yan et al., 2025; Raza et al., 2025b) as well as task-level memories that accumulate over problem-solving attempts (Shinn et al., 2023; Zhang et al., 2025b; Yuen et al., 2025; Xu et al., 2025).”
CL4SE: Benchmarking Context Learning on Software EngineeringPreprint
Hu, H., Zhang, Q., Shang, Y., Xie, G., Fang, C., Chen, Z., & Xiao, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.23047 · Cited by 3
MAGE: Meta-Reinforcement Learning for Language Agents toward Strategic Exploration and ExploitationPreprint
Yang, L., Xu, Z., Xie, M., Gao, J., Shok, Z., Wang, Y., & Wu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.03680 · Cited by 4
“Current approaches to achieve such test-time adaptation primarily rely on prompting through In-Context Learning (ICL) Brown et al. (2020); Shinn et al. (2023) or external memory components Xu et al. (2025); Zhou et al. (2025).”
Proactive Memory for Ad-Hoc Recall over Streaming DialoguesPreprint
Wang, B., Li, J., & Xu, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.04885 · Cited by 3
“( 2) Agentic Memory Systems : MemoRAG (Qian et al., 2025), A-Mem (Xu et al., 2025b), and MemGAS (Xu et al., 2025a).”
AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and ReusePreprint
Zhang, Z., Lu, S., Qian, H., He, D., & Liu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.18000 · Cited by 4
“Con-currently, skill accumulation methods maintain experience through structured memory mechanisms (Tang et al., 2025; Zhou et al., 2025; Xu et al., 2025) or by saving executable tool-level skills, as demonstrated by Voyager (Wang et al., 2023).”
Zhang, W., Zhou, Y., Qu, H., & Li, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.15690 · Cited by 3
“…often lack explicit supply–demand contracts, so coordination devolves into semantic probing) [1, 38, 51], and (iii) persistent-memory and Artifact maintenance (large, evolving stores of skills, traces, and long-term memories accumulate operational debt and require coherent management) [34, 53, 58].”
LLM Agent Memory: A Survey from a Unified Representation–Management PerspectivePreprint
Tang, Z., He, X., Zhao, T., Wei, F., Liu, X., Dong, P., Wang, Q., Li, Q., Wang, H., Chen, R., et al. (2026) · Preprints.org · doi:10.20944/preprints202603.0359.v2 · Cited by 3
Shimomoto, T. (2026) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.19212118 · Cited by 3
Experience as a Compass: Multi-agent RAG with Evolving Orchestration and Agent PromptsPreprint
Li, S., & Ramakrishnan, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.00901 · Cited by 4
“Storing distilled experience as retrievable in-context demonstrations has been shown to effectively shape future behavior (Zhao et al., 2024; Ouyang et al., 2025; Xu et al., 2025b).”
FileGram: Grounding Agent Personalization in File-System Behavioral TracesPreprint
Liu, S., Tian, S., Hu, K., Dong, Y., Yang, Z., Li, B., Yang, J., Loy, C. C., & Liu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.04901 · Cited by 3
ATANT: An Evaluation Framework for AI ContinuityPreprint
Tanguturi, S. S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.06710 · Cited by 3
“A-MEM [7] proposes agentic memory with self-organizing capabilities for LLM agents.”
PRIME: Training Free Proactive Reasoning via Iterative Memory Evolution for User-Centric AgentPreprint
Wang, Z., & Jiang, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.07645 · Cited by 3
When to Forget: A Memory Governance PrimitivePreprint
Simsek, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.12007 · Cited by 3
“…recent systems have begun to incorporate dynamic memory management—including admission control, deletion, and adaptive retrieval (Zhang et al., 2026; Xu et al., 2025)—a simple, general per-memory estimator of post-retrieval outcome association, with convergence guarantees, has not been proposed.”
Can Current Agents Close the Discovery-to-Application Gap? A Case Study in MinecraftPreprint
Ziheng, Z., Tang, H., Zhang, J., Lin, H., Yang, B., Long, Q., Sun, F., Sun, Y., Liang, Y., Wu, Y. N., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.24697 · Cited by 3
“…format outperforms the intuitive “Finding-Explanation-Example” format, likely because delineating conditions under which claims hold enables better understanding of when to apply them—offering insights for memory evolution research (Hu et al., 2025; Xu et al., 2025).”
When Stored Evidence Stops Being Usable: Scale-Conditioned Evaluation of Agent MemoryPreprint
Shao, J., Lu, Y., Zhang, Y., & Luo, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.07313 · Cited by 5
“To support long-horizon interaction, agents are often equipped with memory systems that persist information across sessions and make relevant context available at inference time [Park et al., 2023, Packer et al., 2023, Zhong et al., 2023, Xu et al., 2025, Li et al., 2025].”
Self-citationPRISM: Pareto-Efficient Retrieval over Intent-Aware Structured Memory for Long-Horizon AgentsPreprint
Peng, J., Wan, Z., Liu, W., & Sun, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.12260 · Cited by 3
“MemGPT [16] pages information through a managed context hierarchy under LLM control; A-MEM [25] stores chunks as Zettelkasten-style notes with semantic links; Zep [17] maintains a bi-temporal knowledge graph; MAGMA [7] traverses four orthogonal edge-graphs with a heuristic intent-aware beam search;…”
SAGE: A Self-Evolving Agentic Graph-Memory Engine for Structure-Aware Associative MemoryPreprint
Wang, J., Zhao, H., Pan, G., Wang, X., Wang, Y., Deng, Q., & Zhang, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.12061 · Cited by 4
RealICU: Do LLM Agents Understand Long-Context ICU Data? A Benchmark Beyond Behavior ImitationPreprint
Shen, C., Shen, W., Susetzky, T., Chen, Chen, Li, J., Liu, Y., Zhang, X., Gong, Z., Rueckert, D., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.13542 · Cited by 3
“Retrieval-based systems such as RAG [1, 4] and A-MEM [33] enable selective access over long histories.”
AuthTrace: Diagnosing Evidence Construction in Thematically Dense Single-Author CorporaPreprint
Xiaoqing, W., Li, F., 梁, 铭., & Que, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.25382 · Cited by 3
“Agent-memory systems and benchmarks, including MemGPT, Mem0, A-MEM, LoCoMo, and LongMemEval, study how information is stored, compressed, updated, and retrieved over time (Packer et al., 2024; Chhikara et al., 2025; Xu et al., 2025; Maharana et al., 2024; Wu et al., 2025).”
SAGE: A Novelty Gate for Efficient Memory Evolution in Agentic LLMsPreprint
Wang, S., Brahma, D., & Henao, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.30711 · Cited by 3
“Therefore, any difference be-tween the two methods A-Mem and A-Mem+ SAGE is attributable to SAGE alone.”
InfluentialEywa: Provenance-Grounded Long-Term Memory for AI AgentsPreprint
Joshi, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.30771 · Cited by 3
“…A-MEM, MemoryOS, and graph-based RAG methods further show that external memory, temporal structure, and graph organization can improve long-horizon agents (Chhikara et al., 2025; Su-permemory, 2026; Rasmussen et al., 2025; Xu et al., 2025; Kang et al., 2025; Jiménez Gutiérrez et al., 2024).”
ForecastCompass: Guiding Agentic Forecasting with Adaptive Factor MemoryPreprint
Chang, Y., Du, Y., Cao, Y., Chen, J., & Lin, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.30858 · Cited by 4
“…M EM 0 [Chhikara et al., 2025], which stores and retrieves long-term interaction memories; and reflective or experience-driven memory , represented by R E - FLEXION [Shinn et al., 2024] and A-M EM [Xu et al., 2025], which construct verbal lessons or agent-experience memories from past trajectories.”
MARDoc: A Memory-Aware Refinement Agent Framework for Multimodal Long Document QAPreprint
Chen, K., Liu, C., Peng, Q., Yang, J., Liu, Y., Zhang, X., & Yang, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.05749 · Cited by 6
“For example, Agentic Reasoning (Wu et al., 2025b) and A-Mem (Xu et al., 2025) model tool-calling trajectories as structured reasoning memory, while Reasoning-Bank (Ouyang et al., 2025) extracts transferable tool-usage patterns from historical calls to improve tool-use accuracy and efficiency.”
ConMem: Structured Memory-Guided Adaptation in Training-Free Multi-Agent SystemsPreprint
Tan, Z., (114038), Q. C., Huang, T., Su, X., & (38576), Y. C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.08702 · Cited by 3
“…abstracting reusable procedures from interactions—such as skill libraries [11, 12], procedural memory extraction [13], agentic memory organization [14], MAS memory systems [15, 16], clustered memory organization [17], and dynamic procedural memory pools [18]—these methods reduce redundancy and…”
REAL: A Reasoning-Enhanced Graph Framework for Long-Term Memory Management of LLMsPreprint
Lu, K., Chen, L., Jiang, G., Qin, Z., Liu, Y., & Zhang, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.10694 · Cited by 5
“This separation calls for an external memory repository that persistently stores historical information outside the LLM, indexes it for efficient retrieval, and dynamically supplies relevant subsets back into the context window when needed [12].”
InfluentialWang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.09900 · Cited by 3
AgentSpec: Understanding Embodied Agent Scaffolds Through Controlled CompositionPreprint
Chen, J., Shen, J., Kang, H., Hong, Z., Jiang, Q., Bose, S., Zhang, Y., Leng, L., Vyas, A., Mao, L., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.14674 · Cited by 3
Self-citationGitOfThoughts: Version-Controlled Reasoning and Agent Memory You Can Replay, Diff, and MergePreprint
Shekar, P. C., Abhishek, H. S., & Krishnan, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.14470 · Cited by 3
“Voyager’s skill library [18], Gener-ative Agents’ observation stream [13], MemoryBank [29], MemGPT [12], ExpeL [28], CLIN [10], Mem0 [1], A-MEM [22], and TextGrad [26] all use bespoke side stores.”
Yang, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.15903 · Cited by 3
“…2023) and Letta (Letta contributors, 2024) paginate between main and archival memory; Zep (Rasmussen et al., 2025) maintains a temporal knowledge graph with edge invalidation; HippoRAG (Gutiérrez et al., 2024) uses PageRank over an extracted KG; A-MEM (Xu et al., 2025) applies Zettelkasten linking.”
Memory Shot for Long-Term DialoguePreprint
Peng, C., Xin, H., Sheng, X., Dai, X., Liu, Z., Wang, S., Yan, Y., Chen, Z., Gu, Y., & Yu, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.28338 · Cited by 3
“Mem0 [4] and Zep [29] study salient memory extraction and graph-based organization, while A-MEM [41] focuses on memory evolution and update mechanisms.”
AgRefactor: Self-Evolving Agentic Workflow for HLS Compatibility and PerformancePreprint
邹洋, Ding, Z., Sun, Y., & Cong, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.30949 · Cited by 3
“Relative to general agentic memory systems [13], [14], [15], [16], [17], [18], we create a specialized knowledge format, dedicated interaction patterns, and a novel distance metric, all designed to achieve superior performance on this task.”
The Compliance Trap: Diagnosing How AI Agents Consume Conflicting MemoryPreprint
Chen, Y., Bai, X., & Yuille, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.10608 · Cited by 3
“Helpful memory improves trajectories when grounded; conflicting memory harms when adopted and not recovered.”
Keep It InMind: Benchmarking the Implicit-Association Blind Spot in Agent MemoryPreprint
Li, R., Du, M., Xu, B., & Mao, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.24368 · Cited by 4
“Figure 1 is not a hypothetical: HippoRAG 2 (Gutierrez et al., 2025), A-Mem (Xu et al., 2025), xMemory (Hu et al., 2026), and A-RAG (Du et al., 2026) all fail exactly this task in the benchmark we introduce below.”
InfluentialV-Mem: Modality-Routed Retrieval for Long-Term Multimodal Agentic MemoryPreprint
Kang, D., Jiang, D., Li, Y., Li, G., & Li, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.01543 · Cited by 5
“They distill a conversation down to text, whether summaries, graph nodes, or sentence embeddings, and find evidence by comparing text against text (Xu et al. 2025; Liu et al. 2026b; Jiang et al. 2026a).”
InfluentialMemForest: Efficient Agent Memory Management via EventTree Partitioning and Progressive MergingPreprint
Wang, J., Sun, T., Zhu, J., Zhang, C., Li, S., Liu, X., Wen, Z., Tu, X., Ren, J., Zhang, X., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.08273 · Cited by 3
Safety in Self-Evolving Agents: A SurveyPreprint
Chen, J., Feng, Z., Ma, O., Yan, Y., Lin, R., Zhang, H., Du, L., An, H., Yang, Y., Liu, J., et al. (2026) · arXiv (Cornell University) · doi:10.13140/rg.2.2.23319.74403 · Cited by 4
Decision-Sufficient State Representations: Measuring and Reducing Write-Time RegretPreprint
Shen, B., & Li, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32805 · Cited by 3
“…interaction proceeds (Wang et al., 2023); MemGPT pages information between context and external storage (Packer et al., 2023); Reflexion, generative agents, MemoryBank and A-MEM keep reflections and retrievable memories (Shinn et al., 2023; Park et al., 2023; Zhong et al., 2024; Xu et al., 2025).”
Real-Time Procedural Learning From Experience for AI Agents
Bi, D., Hu, Y., & Nasir, M. N. (2026) · International World Wide Web Conference (WWW) · doi:10.1145/3774905.3795087 · Cited by 2
OmniRouter: Budget and Performance Controllable Multi-LLM Routing
Mei, K., Xu, W., Guo, M., Lin, S., & Zhang, Y. (2025) · arXiv · doi:10.48550/arxiv.2502.20576 · Cited by 2
“This success has spurred the widespread integration of LLMs into modern systems [43; 29; 32; 40; 56; 26; 48].”
Self-citationKim, S., Lee, Y., Kim, S., Kim, H., & Cho, S. (2025) · Findings of the Association for Computational Linguistics: EMNLP 2025 · doi:10.18653/v1/2025.findings-emnlp.1204 · Cited by 2
Shen, Y., Li, K., Zhou, W., & Hu, S. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.370 · Cited by 2
3DLLM-Mem: Long-Term Spatial-Temporal Memory for Embodied 3D Large Language Model
Hu, W., Hong, Y., Wang, Y., Gao, L., Wei, Z., Yao, X., Peng, N., Bitton, Y., Szpektor, I., & Chang, K.-W. (2025) · Advances in Neural Information Processing Systems 38 · doi:10.52202/085713-2278 · Cited by 27
MAGMA: A Multi-Graph based Agentic Memory Architecture for AI Agents
Jiang, D., Li, Y., Li, G., & Li, B. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.1709 · Cited by 2
Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolution
Cao, Z., Deng, J., Yu, L., Zhou, W., Liu, Z., Ding, B., & Zhao, H. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.829 · Cited by 2
HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents
Zhu, J., Li, J., Zhang, C., Liu, J., & Yang, M. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.625 · Cited by 2
Memento: Fine-tuning LLM Agents without Fine-tuning LLMs
Zhou, H., Chen, Y., Guo, S., Yan, X., Lee, K. H., Wang, Z., Lee, K. Y., Zhang, G., Shao, K., Yang, L., et al. (2025) · arXiv · doi:10.48550/arxiv.2508.16153 · Cited by 2
“A-MEM (Xu et al., 2025) maintains memory via a typological network.”
Dive into Claude Code: The Design Space of Today's and Future AI Agent Systems
Liu, J., Zhao, X., Shang, X., & Shen, Z. (2026) · arXiv · doi:10.48550/arxiv.2604.14228 · Cited by 2
Cerebrum (AIOS SDK): A Platform for Agent Development, Deployment, Distribution, and Discovery
Rama, B., Mei, K., & Zhang, Y. (2025) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2025.naacl-demo.14 · Cited by 8
Self-citationWISE-Flow: Workflow-Induced Structured Experience for Self-Evolving Conversational Service Agents
Zhou, Y., Wang, Z., Jk, Wang, H., Koelle, S., Zhu, Z., & Niu, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.08158 · Cited by 3
“Interaction trace memories store fine-grained traces, such as individual tool calls or events (Zhang et al., 2023; Xu et al., 2025).”
ATOD: An Evaluation Framework and Benchmark for Agentic Task-Oriented Dialogue Systems
Zhang, Y., Nayyeri, H., Khaziev, R., Yılmaz, E., Tür, G., Hakkani-Tur, D. Z., & Thadakamalla, H. P. (2026) · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL) · doi:10.18653/2026.aacl-long.109 · Cited by 2
Self-citationAmory: Building Coherent Narrative-Driven Agent Memory through Agentic Reasoning
Zhou, Y., Guo, X., Bayar, B., & Sengamedu, S. H. (2026) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2026.eacl-long.183 · Cited by 8
DeepSieve: Information Sieving via LLM-as-a-Knowledge-Router
Guo, M., Zeng, Q., Zhao, X., Liu, Y., Yu, W., Du, M., Chen, H., & Cheng, W. (2026) · Findings of the Association for Computational Linguistics: EACL · doi:10.18653/v1/2026.findings-eacl.160 · Cited by 11
Towards Effective Experiential Learning: Dual Guidance for Utilization and Internalization
Bai, F., Chen, Z., Hao, C., yang, M., Tao, R., Dai, B., Zhao, W. X., Yang, J., & Xu, H. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.774 · Cited by 2
C law VM: Harness-Managed Virtual Memory for Stateful Tool-Using LLM Agents
Rafique, M. B., & Bindschaedler, L. (2026) · Workshop on Machine Learning and Systems (EuroMLSys) · doi:10.1145/3805621.3807648 · Cited by 8
“A-MEM [44] takes a different approach, organizing long-term memory as a Zettelkasten-like network with dynamic note construction and linking rather than prompt-time paging.”
The agentic AI security adoption matrix: Understanding readiness and resistance across domains
WAIZEL, G. (2026) · Smart Cities and Regional Development (SCRD) Journal · doi:10.25019/hccxvf87 · Cited by 4
“…was developed based on the full corpus of abstracts and studies listed in the references [24, 1, 2, 25, 26, 4, 5, 6, 7, 15], [27, 28, 29, 30, 31, 32, 33, 34], [35, 36, 37, 38, 39, 40, 41, 42, 16], [17, 43, 9, 11, 10, 12, 14, 13], [18, 19, 9, 44, 45, 46, 47, 48, 49],[50, 23, 20, 22, 21, 51].”
TA-Mem: Tool-Augmented Autonomous Memory Retrieval for LLM in Long-Term Conversational QA
Yuan, M., Liu, J., Yang, J., Li, X., Yan, W., Wu, Y., & Liang, P. (2026) · International Conference on Advanced Algorithms and Control Engineering (ICAACE) · doi:10.1109/icaace69793.2026.11509181 · Cited by 19
“A-Mem [10] uses agentic design for memory generation, in which the LLM processes the context into multiple notes and empowers it to dynamically decide when and which memory notes to update.”
R 2 -Searcher: Calibrating Retrieval and Reasoning Boundaries for Agentic Search
Zhang, S., Li, J., Zhang, W., Qian, X., Wang, Y., Zhang, Y., Wang, M., Liu, Y., & Zhao, X. (2026) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3799682.3840698 · Cited by 2
Self-citationMemoBrain: Executive Memory as an Agentic Brain for Reasoning
Qian, H., Cao, Z., & Liu, Z. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.127 · Cited by 3
MemSearcher: Iterative Memory Integration for Search Agent via End-to-End Reinforcement Learning
Yuan, Q., Lou, J., Li, Z., Chen, J., Lu, Y., Lin, H., Sun, L., Zhang, D., & Han, X. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.736 · Cited by 4
“(3) Structured memory (Zeng et al., 2024; Zhang et al., 2025a; Chhikara et al., 2025) organizes information into structured representations, such as the knowledge graphs utilized in Zep (Rasmussen et al., 2025) and the atomic memory units in A-MEM (Xu et al., 2025).”
ARC: Active and Reflection-driven Context Management for Long-Horizon Information Seeking Agents
Yao, Y., Huang, S., Dai, E., Tan, Z., Duan, Z., Jia, S., Jiang, Y., & Yang, T. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.930 · Cited by 7
A Survey of Large Language Model-Based Search Agents
Xi, Y., Lin, J., Xiao, Y., Zhou, Z., Shan, R., Gao, T., Zhu, J., Liu, W., Yu, Y. H., & Zhang, W. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.374 · Cited by 10
“Agentic search can extract queries from complex and ambiguous user intentions and then conduct deep searches within the agent’s memory to retrieve highly pertinent information (Xu et al., 2025b; Ocker et al., 2025; Tan et al., 2024b).”
Sasanelli, L. D., Barbieri, U., & Fuccio, R. D. (2025) · Giornale Italiano di Educazione alla Salute, Sport e Didattica Inclusiva · doi:10.32043/gsd.v9i2.1465 · Cited by 2
Task-Aligned Tool Recommendation for Large Language ModelsPreprint
Gao, H., & Zhang, Y. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2411.09613 · Cited by 2
“Memory recommendation (Borgeaud et al., 2022; Gao and Zhang, 2024a; Xu et al., 2025) facilitates the retrieval of relevant past experiences or interactions, improving continuity, consistency, and long-term context in multi-turn conversations.”
Self-citationPlanet as a Brain: Towards Internet of AgentSites based on AIOS ServerPreprint
Zhang, X., & Zhang, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.14411 · Cited by 2
“The rapid progress of large language models (LLMs) has led to the emergence of autonomous agents capable of planning, reasoning, and interacting with humans and other agents in structured environments [17, 22, 21, 18].”
InfluentialSelf-citationCerebrum (AIOS SDK): A Platform for Agent Development, Deployment, Distribution, and DiscoveryPreprint
Rama, B., Mei, K., & Zhang, Y. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2503.11444 · Cited by 2
“Acting as core coordinators, agents are envisioned as intelligent entities that can perceive their surroundings, build memories (Xu et al., 2025a; Wang et al., 2024b), devise plans, and autonomously carry out actions to fulfill tasks (Wang et al., 2024a; Deng et al., 2024; Shi et al., 2025; Xu et…”
Self-citationCodeMEM: AST-Guided Adaptive Memory for Repository-Level Iterative Code GenerationPreprint
Wang, P., Zhang, L., Liu, F., Tao, C., & Zhu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.02868 · Cited by 2
“Baselines like MemGPT and A-Mem, which performed similarly to FC on CodeIF-Bench, show weaker performance here, highlighting their limited generalizability.”
InfluentialZhang, X., Yang, K., Li, H., Li, C., Wei, Q., & Ananiadou, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.03417 · Cited by 2
“…for episodic memory) (Kim et al., 2025), Mem0 / Mem0g (production-oriented long-term memory with an optional graph variant) (Chhikara et al., 2025), A-Mem (agentic memory with dynamic indexing and linking) (Xu et al., 2025), and MemGen (generative latent memory tokens) (Zhang et al., 2025).”
InfluentialDep-Search: Learning Dependency-Aware Reasoning Traces with Persistent MemoryPreprint
Liu, Y., Peng, X., Yan, Z., Shen, Y., Xu, W., Huang, Y., Wang, X., Cao, J., Yin, J., & Zhang, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2601.18771 · Cited by 2
“A growing line of work studies how LLM agents can maintain and exploit persistent memory to improve long-term coherence, personalization, and knowledge reuse [19, 21, 37].”
Self-citationYang, W., Liu, Z., Li, Y., Yan, B., Li, L., He, M., Wojtczak, D., Zheng, Y., & Shi, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.06902 · Cited by 2
“To further evaluate the effectiveness of mapping experiences into an entity-agnostic abstract skill space, we compare our framework against recent state-of-the-art (SOTA) agentic memory paradigms, including ExpeL [24] and A-Mem [25].”
Sun, Y., Yu, H., Jiang, W., Yu, X., Zhan, D., Wang, L., Ren, S., Sun, Y., & Zhu, T. (2026) · Preprints.org · doi:10.20944/preprints202602.1655.v2 · Cited by 2
Self-Evolving LLM Memory Extraction Across Heterogeneous TasksPreprint
Yang, Y., Liu, T., Zhu, W., Shi, T., Song, L., & Jia, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.11610 · Cited by 2
Retrieval Is Not Enough: Why Organizational AI Needs Epistemic InfrastructurePreprint
Bottino, F., Ferrero, C., Dosio, N., & Beneventano, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.11759 · Cited by 2
“A controlled pilot comparison ( n =10 response pairs) with full confound treatment, plus three formal design properties distinguishing KGE from static epistemic labeling.”
Bian Que: An Agentic Framework with Flexible Skill Arrangement for Online System OperationsPreprint
Liu, B., Qian, Z., zhao, Y., Jiang, X., Liang, Z., Ma, Y., Zhuang, J., Chen, B., Yang, S., Wan, H., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.26805 · Cited by 2
“On the knowledge side, prior work spans RAG [30] and its variants [31, 32], memory-augmented architectures [33, 34, 35, 36], and self-refinement methods [37, 38].”
Self-citationEviMem: Evidence-Gap-Driven Iterative Retrieval for Long-Term Conversational MemoryPreprint
Li, Y., He, Y., Zhang, Z., & Gong, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.27695 · Cited by 2
“Persistent memory systems such as Mem-oryBank (Zhong et al., 2023) and Generative Agents (Park et al., 2023) support long-term consistency, while recent work adds structural priors: A-Mem (Xu et al., 2025) for agent-controlled memory selection, MAGMA (Jiang et al., 2026) for multi-graph reasoning, and TReMu (Ge et al.…”
ShadowMerge: A Novel Poisoning Attack on Graph-Based Agent Memory via Relation-Channel ConflictsPreprint
Luo, Y., Kang, Z., Ji, T., Liu, X., Liu, Y., Li, S., & Peng, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.09033 · Cited by 2
MAGE: Multi-Agent Self-Evolution with Co-Evolutionary Knowledge GraphsPreprint
Yang, R., Li, Z., Xue, H., Razzak, I., & Salim, F. D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.10064 · Cited by 3
“Flat episodic memory [21, 32, 23, 2] stores trajectories or principles, but provides little structure for curriculum selection or dependency-aware reuse.”
MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent ReasoningPreprint
Liu, Y., Ye, Z., Sun, Y., Zhu, M., Xiao, J., Han, Z., Qi, G. U., Cai, X., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.13037 · Cited by 2
“…to optimize this cycle: improving reasoning capability through expert trajectories, parameter optimization, or experience replay [16, 18, 23]; and enhancing memory systems through external trajectory storage or distilled knowledge retrieval to augment decision-making context [8, 26, 34, 37].”
ClawForge: Generating Executable Interactive Benchmarks for Command-Line AgentsPreprint
Lai, Y., Xia, P., Ji, H., Xiong, K., Zeng, K., Liu, J., Wu, F., Zhong, J., Zheng, Z., Xie, C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.14133 · Cited by 2
Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM DeliberationPreprint
Yang, J. C., Flechtner, M., Dailisan, D. N., & Bakker, M. A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.15343 · Cited by 2
“…can be retained, reflected on, or retrieved: Generative Agents (Park et al., 2023), episodic memory banks (Zhong et al., 2023), reflection systems (Xu et al., 2025; Tan et al., 2025), graph memory (Gutiérrez et al., 2024; Kang et al., 2025; Huang et al., 2025), large-scale simulations (Piao et…”
EXG: Self-Evolving Agents with Experience GraphsPreprint
Jin, Y., Zhang, S., Wang, H., Qin, L., Zhang, Y., & Zhang, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.17721 · Cited by 3
“…a broader and more mature research direction, encompassing methods that model persistent context or knowledge for long-term coherence [5, 26, 28, 32, 38], treat memory as a system-level or control component governing memory organization and lifecycle [7, 18, 41, 42], and conduct efficiency…”
Self-citationHe, X., Lin, Z., Liu, S., Wu, J., Xie, Q., Zhou, W., & Xiao, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.23067 · Cited by 2
“A-Mem [2] introduces dynamic agentic memory with structured entries.”
Connecting the Dots: Benchmarking Reflective Memory in Long-Horizon DialoguePreprint
Lin, J., Wang, B., Wang, Z., Jin, Z., Qiao, W., Li, J., & Xu, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.01223 · Cited by 2
“② training-free memory methods : MemGPT [Packer et al., 2023], Mem0 [Chhikara et al., 2025], MemoryOS [Kang et al., 2025], LangMem [LangChain, 2025], A-Mem [Xu et al., 2025], LightMem [Fang et al., 2025], and GAM [Yan et al., 2025b].”
InfluentialLin, H., Pao, H., Zhan, S., & Zheng, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.00288 · Cited by 2
Min-Cost Flow Routing for Evidence Assembly in Long Multimodal DocumentsPreprint
Mehrish, A., & Vascon, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.07235 · Cited by 2
“G 2 -Reader (Du et al., 2026) maintains two evolving graphs consolidated through VLM message passing, an approach related to graph memories for long-horizon agents (Chhikara et al., 2025; Xu et al., 2025b; Rasmussen et al., 2025).”
Cai, Y., Guo, X., Huang, X., Du, J., Huang, C., Huang, W., Ma, W., Hu, Y., Zeng, A., Tang, J., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.08656 · Cited by 2
Light-Omni: Reflex over Reasoning in Agentic Video Understanding with Long-Term Memory
Nie, C., Wei, J., Feng, J., Fu, C., & Shan, C. (2026) · IEEE Transactions on Multimedia · doi:10.48550/arxiv.2607.05511 · Cited by 2
“In practice, these Polaroids correspond to external databases or evolving memory banks, accessed via retrieval mechanisms to compensate for missing context [17, 50, 59].”
InfluentialWhat's Missing in Autonomous Research? A Systematization of Systems, Benchmarks, and VerificationPreprint
Ren, X., Sun, Y., Yi, C., Zhang, K., Guo, J., Du, J., & Yang, H. (2026) · Research Square · doi:10.21203/rs.3.rs-10060294/v1 · Cited by 4
Bouzidi, H., Mkpong, M. E., & Faruque, M. A. A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.15657 · Cited by 2
“These agents rely on multimodal persistent long-term memory: visual and textual content from past interactions are encoded into a shared embedding space, persisted across sessions, and retrieved at query time [86, 54, 7, 71, 43, 73, 57, 45, 18].”
Scaling LLM Agent Tool Access: The MCP Gateway
Li, M., Song, E., Zuo, Y., Liu, X., Wen, R., Fu, Q., Antichi, G., He, J., Tie, J., Shao, Z., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.15593 · Cited by 3
MemChain: Learning Interpretable Memory Traces for Memory-Augmented LLM AgentsPreprint
Ma, Y., Tu, S., Zhang, Q., Li, D., Li, L., & Zhao, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.24097 · Cited by 3
“…memory-management methods optimize construction or management behavior us-ing supervised or reinforcement signals (Chhikara et al., 2025; Xu et al., 2025; Wei et al., 2026; Liu et al., 2026a; Zhang et al., 2026a; Xu et al., 2026; Shen et al., 2026; Yan et al., 2025; Wang et al., 2026).”
InfluentialFu, S., Tao, Y., Jin, L., Zhao, J., Shi, Q., Pan, T., Li, H., Wang, C., Wu, J., & Huo, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.02143 · Cited by 2
Continual Learning in TransitionPreprint
Hou, Z., Zhang, D., Feng, T., Wang, L., Li, W., Hao, X., An, H., Fang, J., Ma, H., Xu, Z., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.06216 · Cited by 3
Self-Correcting Long-Horizon Search Agents via Tree-Structured MemoryPreprint
Yang, A., Guo, Q., Huang, Z., Chen, Y., Qian, S., & Cao, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.10676 · Cited by 2
What to Remember, What to Reveal: Privacy-Aware Memory for Conversational AgentsPreprint
Wang, W., Si, W., Xu, X., & Xu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.16551 · Cited by 2
“Vector-based memory stores conversations or extracted facts as dense embeddings for semantic retrieval, as in MemoryBank [2] and A-MEM [12]; A-MEM further augments memory notes with attributes and dynamic links.”
HyMem: Hierarchical Context Management for Long-Horizon Agents via Information IsolationPreprint
Wang, X., Xiao, J., Cui, S., Zhang, H., Wang, Y., Zhang, Q., & Xu, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.15703 · Cited by 2
“On Browsecomp-plus, HyMem reaches an average Pass@1 of 61.3, outperforming ReAct by 4.7 percentage points and substantially exceeding memory-oriented or modular baselines such as ReSum, A-MEM, and THREAD.”
InfluentialFresh Memory, Stale Plans: Derivation Currency for Distributed LLM-Agent MemoryPreprint
Chen, E., Wang, S., & Brinton, C. G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.03340 · Cited by 2
“Registers, replication, and consistency policies offer different guarantees and costs for exposing shared state [20, 21, 22, 23, 24, 25, 26], while agent frameworks structure interaction [27, 28] and memory systems support retrieval and sharing [29, 30, 31, 7, 8].”
Interchangeable in Role, Not in History: Replacing Members of Formed LLM TeamsPreprint
Gao, J.-X., Li, R.-Z., Yu, T.-Y., Ren, L.-W., Chen, B.-H., & Wang, Z.-N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.05279 · Cited by 1
“Reflection and memory mechanisms let an agent carry what it learned from one episode to the next without touching its weights [26, 32, 49], and a growing line of work pools such memory across trials or shares it between agents [7, 52].”
Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory PortabilityPreprint
Goyal, A., & Ray, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.05339 · Cited by 4
AIM: A Privacy-Aware Interoperable Memory Framework for Multi-Agent Multi-User LLM Systems
Johnson, Z., Kumankumah, N. B., Chatterjee, S., Sathyamurthi, T., Chen, M., Li, X. A., Wang, X., Gelchie, E. M., Lin, J., Hasan, S. A., et al. (2026) · arXiv
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.