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 801–900 of 1,030
CLAG: Adaptive Memory Organization via Agent-Driven Clustering for Small Language Model Agents
Roh, T., Jang, W., Jung, J., & Kang, J. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.824 · Cited by 2
Guo, J., Su, Y., Ge, Y., Guan, S., Yu, L., Guo, J., & Cheng, X. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.2097
“…2024c; Qian et al., 2025), multi-agent collaboration (Zhang et al., 2024b; Zhao et al., 2024a), memory augmentation (Zhou et al., 2023; Zhong et al., 2024; Wang et al., 2024; Chhikara et al., 2025; Xu et al., 2025), and context compression (Lee et al., 2024; Zhang et al., 2024a; Hu et al., 2025).”
RSMeM: Knowledge-Enhanced Memory Evolution for Remote Sensing Agents with Systematic Evaluation
Wu, B., Zhang, Y., Guo, Z., Liu, T., Qian, C., Lu, Y., Du, X., Li, Y., Zhang, Y., Chen, C., et al. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1519 · Cited by 1
“Memory enables language agents to retain and reuse information across interactions, substantially improving adaptability and task performance (Fang et al., 2025; Xu et al., 2025b).”
Self-citationFrom Personal to Collective: On the Role of Local and Global Knowledge in LLM Personalization
Wang, Z., Wu, J., Tan, Z., Li, B., Zhong, X., Liu, Z., & Zeng, Q. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.1478
“Beyond retrieval, several works de-sign long-term knowledge systems (Packer et al., 2023; Wang et al., 2023a; Wei et al., 2025; Xu et al., 2025) to better capture and organize user-specific context over extended interaction sessions.”
AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse
Zhang, Z., Lu, S., Qian, H., He, D., & Liu, Z. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-demo.81 · Cited by 1
Graph-Assisted Large Language Models: A Perspective on Mitigating Intrinsic Limitations
Luo, H., Wang, F., Zhang, W., Zhang, X., Zhiwei, Z., Zhao, T., Lin, M., Zhang, J., Liu, H., Tang, X., et al. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.945 · Cited by 4
“Planar graphs directly connect raw interaction records and enrich representations via neighborhood aggregation but are less explored (e.g., A-mem (Xu et al., 2025)).”
Self-citationAMA: Adaptive Memory via Multi-Agent Collaboration
Huang, W., Wang, Z., Lin, H., Wang, S., Xu, B., Li, Q., Zhu, B., Yang, L., & Qin, C. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.152 · Cited by 6
A Dual-Phase Self-Evolution Framework for Large Language Models
Sun, H., Zhang, Z., & Zeng, S. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.37
“A-MEM (Xu et al., 2025) constructs an evolving knowledge graph, improving organization at the cost of structural complexity.”
InfluentialThe Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
Chen, Z., Dong, S., Shi, C., Wang, P., Wang, S., Shen, C., & Li, J. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.677
Gao, W., Fan, X., Yeh, C. M., Sun, J., Chen, Y., Pan, M., Das, M., & Liu, Y. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.619
“While there exist general frameworks for memory and feedback mechanisms (Lu et al., 2025; Chhikara et al., 2025; Xu et al., 2025), F2R is designed specifically for LLM-assisted molecular optimization.”
Episodic-to-Semantic Consolidation Without Identity DriftPreprint
Qin, X., Luan, S., Yang, C., & Li, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.01988
Mastermind: Strategy-grounded Learning for Repository-Scale Vulnerability ReproductionPreprint
Du, M., Tuan, L. A., Wu, T., Liu, R., Guo, Z., Huang, D., & Ng, S.-K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.01764
“Cross-run experience, no trainable planner : A-Mem [10] stores agent memories, while AlphaEvolve [11] maintains and evolves prior candidates; both leave the central planner largely frozen.”
Memory-Orchestrated Semantic System (MOSS): An Auditable Agentic Memory ArchitecturePreprint
Lacasse, S., HATIER, J., & Baker, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.04391
“From one side, agentic memory is now a named category: A-MEM [Xu et al., 2025] organizes memories as an evolving Zettelkasten-style network under agent control; AgeMem [Yu et al., 2026] exposes memory operations as tool actions within the agent’s own policy; and “beyond RAG for agent memory” is now…”
Akashic: A Low-Overhead LLM Inference Service with MemAttentionPreprint
Liu, Y., Luo, Z., Jin, H., He, R., Hong, C., Wang, Z., Liu, Y., Gu, Y., Wu, C. E., & Hu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.05708
Twelve Voices, One Decision: Evaluating a Persona-Memory FOMC Simulator Across Seventeen MeetingsPreprint
Suh, J. Y., & Kim, S.-H. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6932158
Task-Adaptive Retrieval over Agentic Multi-Modal Web Histories via Learned Graph Memory
Forouzandeh, S., Berahmand, K., & Jalili, M. (2026) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3805712.3809978
“Mem0[4] and A-MEM [25] extract salient experiences and retrieve them via embedding similarity, assuming relevance is static—an assumption violated in long-horizon navigation.”
ProEchoMem: Enhancing Long Video Understanding via Multi-Trace Probe-Echo Memory
Xu, D., Chen, Y. S., Wang, W., Jia, P., Zhang, C., Wang, M., Wang, Y., Qiang, J., Wei, X., Yin, H., et al. (2026) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3805712.3809599 · Cited by 1
Lee, G.-H., & Park, J. Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.24765
“A-MEM [38] organizes memories via linking; Memoria [39] combines summarization with knowledge graphs; Mem0 [40] provides production-grade continuity; and MemoryOS [43] proposes an OS abstraction for agent memory.”
Ding, C., & Zhu, W. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3520-9_15
Ma, J., Wang, C., Zhang, Y., Wang, Q., Yang, W., & Zhao, Z. (2026) · Lecture notes in computer science · doi:10.1007/978-981-92-3403-5_9
Self-citationLin, Y., Cheng, Y.-C., & Lee, C.-H. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-33381-0_31
Using a Multi-agent AI System for Emotion-Aware Music Recommendation
Cheng, Y.-C., Yang, H.-C., Lin, Y., & Lee, C.-H. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-33381-0_32
Mechanistic Attention Guidance for Agent Memory RefinementPreprint
Hong, Y., Qiu, H., Wang, Y., & Yao, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.17621
“Prior work has explored diverse memory forms, including trajectory exemplars, contextual memories for planning, structured long-term memory, and hierarchical memory management [44, 14, 20, 12, 35, 13].”
ZifaMem: Structured Memory for Persona, Preference, and Emotional Continuity in AI CompanionsPreprint
Fang, J., Xu, G., Cui, Y., Yang, X., & Hua, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.17564
“42 [+0.25,+0.60], and its affect-tagged consolidation is the mechanism the contrastive evidence favors—multi-turn emotional threads beat single-turn snapshots, while injected affect blocks and latent state machines do not earn their place.”
InfluentialRECAP: Feedback-Driven Streaming Semantic User Profiles for Short-Video Recommendation
Zhao, Z., Zhou, X. J., Lv, X., Li, Y., He, C., Liu, Z., Shen, J., Liu, Y., Li, H., Zhang, C., et al. (2026) · ACM Conference on Recommender Systems (RecSys) · doi:10.1145/3773078.3831930
Co-Evolving Graph and Text Memory for Training-Free Multi-Hop Question AnsweringPreprint
Man, H., & Nguyen, T. H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.23278
“SE-Search (Li et al., 2026a) purifies evidence through a Think-Search-Memorize loop, MemSearch-o1 (Zhang et al., 2026) grows reasoning-aligned memory paths, and A-MEM (Xu et al., 2025) revises textual notes retroactively.”
RSMeM: Knowledge-Enhanced Memory Evolution for Remote Sensing Agents with Systematic EvaluationPreprint
Wu, B., Zhang, Y., Guo, Z., Liu, T., Qian, C., Lu, Y., Du, X., Li, Y., Zhang, Y., Chen, C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.24772 · Cited by 2
“Memory enables language agents to retain and reuse information across interactions, substantially improving adaptability and task performance (Fang et al., 2025; Xu et al., 2025b).”
Self-citationLiu, D., He, H., Wu, X., & Li, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.28103
“…interaction histories (Park et al. 2023; Zhong et al. 2024), distilled reflections (Shinn et al. 2023; Zhao et al. 2024), and reusable skills or workflows (Wang et al. 2023, 2024); recent systems also organize such memories automatically at production scale (Xu et al. 2025; Chhikara et al. 2025).”
ViSAGE: Constructing Self-Correcting Memories for Long-Form Video Understanding
Zhao, X., Chen, E., Zhang, Y., Liu, C., Cheng, G., Wang, N., & Xu, Y. (2026) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3767308.3835852
Self-citationDP-MemView: A Memory Interface for Attribute-Level Transcript Privacy in Long-Term LLM AgentsPreprint
Kim, J. W., Min, B., Edemacu, K., Choi, Y., Cho, S.-H., & Jang, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.03130
Yifan, Z. H. O. U., Duan, H., Jianhua, W. A. N. G., Hongli, Z. H. A. O., 刘式达, & Xingyan, T. A. N. (2026) · Shuili Xuebao · doi:10.3724/j.slxb.20250475
Closing Reasoning Gaps in Clinical Agents with Differential Reasoning Learning
Liu, J., Jiang, Y., Krishnan, R., Padman, R., Zhang, Y., & Bian, J. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3818958
“Because clinical questions are typically multi-factorial involving findings, intermediate hypotheses, and recommended actions, we represent each reasoning trace as a reasoning graph: a directed acyclic graph (DAG) that makes the underlying relational structure explicit.”
Self-citationValutazione e intelligenza artificiale. Prospettive didattiche, etiche e metodologiche
L. Giannandrea, F. G., Giannandrea, L., Gratani, F., & Capolla, L. M. (2026) · U-PAD Unimc - Open Digital Publications (University of Macerata)
Unified Agent: Managing Interactions across DevicesPreprint
Liu, X., Li, R. B., Wei, S., Lin, X., & Nguyen, T. Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.05729
“Agent memory systems store and retrieve extracted facts, evolving notes, temporal graphs, or user profiles (Chhikara et al., 2025; Xu et al., 2025; Packer et al., 2023; Rasmussen et al., 2025; Shu et al., 2026; Shaikh et al., 2025; Yao et al., 2025; Rezazadeh et al., 2025).”
When Do Prompt-Side Agent Playbooks Transfer?Accuracy, Cost, and Runtime Shift in Agent Deployment
Lin, W., Zhang, X., & Sun, L. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-industry.251
Shared Substrate: A Discipline for Sustained Human–AI Coupling on Complex ProblemsPreprint
Roshka, O. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7218019
Liu, Y., Wu, W., Zhao, J., & Guo, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.09507
Muscle Memory for Agents: Compile not Merely RetrievePreprint
Omran, P. G., Lanka, S., Zhang, Q., & Dixit, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.08995
“…Lewis et al., 2020), virtual memory hierarchies (MemGPT; Packer et al., 2023), verbal self-reflections (Reflexion; Shinn et al., 2023), Zettelkasten-style reasoning memory (A-Mem; Xu et al., 2025), or callable skill libraries (Voyager; Wang et al., 2024a) and agentic skills (Jiang et al., 2026).”
LatticeMind: A Conflict-Aware Memory Primitive for Multi-Agent Systems
Zhou, H., Zhang, L., Fan, Y., He, T., Chen, S., Geng, H., Torr, P. H. S., & Yin, Z. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1564
Li, X., Li, H., Zhou, Y., Pan, Q., Wang, H., Liu, Y., Zhang, M., & Ai, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.09109
“A-Mem [31] organizes memories as an evolving network of notes.”
InfluentialTang, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.11654
Beyond Retrieval: Query-Conditioned Reuse of Long-Horizon Agent TrajectoriesPreprint
Li, Y., Wang, H., Zhang, L., Huang, M., Zhang, X., Liu, J., Yan, H., & Xu, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.12847
“Memory systems study storage, indexing, updating, and context delivery, including MemGPT, MemoryBank, Hip-poRAG, Mem0, and A-MEM (Packer et al. 2023; Zhong et al. 2024; Guti´errez et al. 2024; Chhikara et al. 2025; Xu et al. 2025).”
ERSkill: Evolving for Skill-Guided Adaptive Memory RetrievalPreprint
Chen, H., Zhang, L., Li, Z., Xue, L., & Zhu, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.12720
“To separate memory representation from retrieval policy, we replace the original RAG memory of Dynamic Cheatsheet, ReasoningBank, and GEPA with ERSkill’s multi-primitive substrate while keeping their evolution procedures unchanged.”
Turenne, N., Sklab, Y., Chenin, E., & Zucker, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.14587 · Cited by 1
Mental Model Management: An Operator-Based Framework for LLM MemoryPreprint
Krämer, O. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.15451
QUMem: Personalized Memory for Query-Conditioned User-State Inference in LLM AgentsPreprint
Wang, H., Li, Y., Zhang, L., Li, P., Che, X., Zhang, X., & Yang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.16168
Memory Canonicalization: A Framework and Benchmark for Cross-Model Drift in Persistent LLM MemoryPreprint
Vadnere, A., & Lonarkar, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.05124
Shukla, A. (2026) · International Journal of Computer Applications · doi:10.5120/ijca5819570ca300
Training a Knowledge Base: Supervised Structure Learning for Agent-Curated Document StoresPreprint
Pan, Y., & Yu, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.21829
“A parallel line accumulates experience across episodes — external memory paging [22], linked long-term note stores [12], [23], reflection over an event stream [24], and distilled feedback or reusable workflows [7], [25], [26].”
ECHO: An Auditable Dual-Path Memory System for Long-Horizon AgentsPreprint
Qian, Y., Miao, H., Guo, B., Jiang, T., Zhao, S., Le, T., Li, L., & Liu, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.21755
“Generative Agents stores observations, reflections, and plans [15]; MemGPT treats long-lived context as a virtual-memory management problem [16]; A-MEM organizes evolving notes with links [17]; Mem0 extracts and updates salient memories [14].”
From Inertia to Objectivity: Improving Deep Research Agents with Noise Isolation
Zhang, X., Zhang, Z., Fu, Z., Lin, B., & Wang, W. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1860 · Cited by 1
Jiang, B., Yuan, Y., Hao, Z., Liu, Y., Shen, M., Chen, S., Wornell, G. W., Callison-Burch, C., Ungar, L., Roth, D., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.21381
“A-MEM explores agentic memory organization through dynamically linked notes rather than fixed memory schemas (Xu et al., 2025).”
StateTune: Transforming LLM-Assisted EDA Flow Tuning into a Stateful, Closed-Loop ProcessPreprint
Li, K., Yao, S., Zheng, S., & Wang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.23601
“Reflexion [17] introduced verbal reinforcement learning via trajectory-level reflection; A-MEM [22] organizes knowledge notes via Zettelkasten-style indexing; MemoryOS [11] designs a hierarchical short/mid/long-term memory; and CFGM [23] grounds memories at multiple granularities for planning.”
Brain Cognition-Inspired Adaptable LLM Agents for Continuous Knowledge Reuse Across Tasks in IoT
Chen, T., Sui, M., Zhang, B., Cao, S., Zhang, W., Qi, L., & Li, X. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3729453
Zhang, Z., & Zhang, W. (2026) · Asia Pacific Economic and Management Review · doi:10.62177/apemr.v3i4.1704
“The mismatch is res olved by context management [16][17][18][19][20][21][22][23][24][29][30][31][32][33][37] —truncation, summarization, selective eviction, or off-loading to external memory [1][2][5][34][35][36][47][49][50] —all sharing one unexamined assumption: what survives compression is what…”
Compact-Memory LLM Agents via Online Max-Member Clustering and Atom-Aware PackingPreprint
Geng, J., Wang, J., & Yuan, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.04915
“When published agentic-memory methods are run at the same ∼ 4 k-token budget under the authors’ original algorithms, MemGPT [15], MemoryBank [21], and A-MEM [18] reach 0 .”
Shanmugavadivelu, P. (2026) · International Journal of Artificial Intelligence and Machine Learning · doi:10.51483/ijaiml.6.9s.2026.1500-1512
Closing the Consistency Gap: Self-Evolving Agents That Learn to Stay on CoursePreprint
Duesterwald, E., Elder, B., Ngweta, L., Ubaru, S., & Zimoń, M. J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.08832
Zhu, L., Lin, Y., Wang, Y., Liu, Y., Li, H. "., & Chen, Y. (2026) · International Workshop on Human-Centered Sensing, Modeling, and Intelligent Systems · doi:10.1145/3842436.3843817
“Modern agent stacks split memory into two roles: the KV cache accelerates inference over the active context, while episodic or structured stores (MemGPT [18], Mem0 [4], A-MEM [23]) hold durable facts.”
MEMO: Multimodal Evidence Memory Organization for Long-Horizon LLM AgentsPreprint
Gao, X., Wang, J., Ruan, J., Cao, G., Liu, T., & Fu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.07471
“Graph and event memory systems such as MAGMA, SYNAPSE, ES-Mem, CAM, A-MEM, and Amory show that temporal, causal, entity, event, and narrative relations can support long-horizon reasoning (Jiang et al. 2026a,b; Zou et al. 2026; Li et al. 2025; Xu et al. 2025; Zhou et al. 2026).”
CreaMem: A Scene-Aware Memory Architecture for Personalized Agents
Sun, Q., Que, Y., He, B., Guo, J., Yang, D., Situ, W., & Ma, C. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.841
Wang, Y., Lin, X., Xu, Y., Gao, S., Yang, R., Huang, K., Wang, Y., Wu, J., Liu, W., Luan, J., et al. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1211
Kernel-Managed Shared Memory for System-Wide PersonalizationPreprint
Lum, R., & Zhang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.10144
“…improve memory’s structure and adaptability – Mem0 [Chhikara et al., 2025], Zep [Rasmussen et al., 2025], LangMem [LangChain Team, 2025], and A-MEM [Xu et al., 2025] – but memory remains an application-layer component in all of them; we instead treat access, filtering, formatting, and injection as…”
Self-citationBalasubramanian, A., Puvvadi, M., & Garg, S. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-25620-1_34
Governed Semantic-State Reinstantiation Across Fresh Reasoning Instances
Webb, N. (n.d.) · PhilPapers (PhilPapers Foundation)
Tessera: Lossless-First, Vendor-Free Session Compression for Long-Horizon Agent HarnessesPreprint
Mosafer, M. (2026) · Research Square · doi:10.21203/rs.3.rs-11000146/v1
From Human-Graph to Agent-Graph Interaction: State of the Art and Future Directions
Bhowmick, S. S., Khan, A., & Choi, B. (2026) · Proceedings of the VLDB Endowment · doi:10.14778/3827998.3828143
“Recent approaches introduce adaptive graph architectures that incrementally incorporate new experiences while preserving consistency [67].”
From Retrieval to Action: A Survey of Agentic Retrieval-augmented GenerationPreprint
Anwar, M. R. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7428138
Memory-Skill Isomorphism: One Skill Carrier, Two Native UsesPreprint
Ruiyuan, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.16669
Personalizing Long-Term Interactions with LLM-Based Agents
Westhäußer, R. (2026) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2026/953
“…maintain consistency, including a short-term memory (STM) for the immediate conversation flow, summaries of prior conversations to provide topical continuity, and a long-term memory (LTM) for retaining information beyond the context window based on the concept proposed by A-Mem [Xu et al. , 2025].”
WFM: Wiki Foundation Model for Complex Agentic ReasoningPreprint
Dong, J., Luo, L., Zhang, S., Chen, G., Guo, T., Yu, Y., Tao, R., Guo, T., Zhang, Q., An, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.18182
AutoViewMem: Self-Configuring Orthogonal Views for Conversational Long-Term MemoryPreprint
Cao, Z., Qu, X., Gu, Z., Chen, X., Yuan, D., Hou, Y., Zhou, S., Gong, J., Huang, J., & Mei, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.21940
“Building on this, a growing body of work equips LLM agents with long-term memory through memory paging, external modules, time-aware updating, hierarchical storage, and scalable consolidation [22, 25, 29, 6, 12, 27, 7, 28, 16, 4, 3, 5].”
PSD: Pseudo Self-Distillation of Memory Representation Capabilities for LLM AgentsPreprint
Suhail, P., Xia, M., Zhang, X., Das, M., Bansal, C., & Rajmohan, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.23449
MoM: Memory of MemoryPreprint
Qin, B., & Lu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.25054
“…systems [10, 7] manage placement —what stays in context under a size limit—not which value holds; so do structural, agentic-linking, hierarchical, and preference-aware memories [20, 18, 13, 14] and graph-RAG archives (HippoRAG-v2) [3], which organize evidence richly but resolve conflicts per query.”
SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party DialoguePreprint
Zheng, H., Tang, T., Chen, Y., Wang, W., & Wu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.26780
MACBT: A Multi-Agent Cognitive Behavioral Therapy Decision Support System with Longitudinal MemoryPreprint
Jiang, D., Zhang, S., Liao, W., Zhang, J., Yu, C., Liao, H., & Yuan, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.30939
“Long-term memory mechanisms such as MemoryBank [13], generative agents [22], and A-MEM [23] enable personalized, cross-session interactions.”
Highlight-Then-Summarize: Learning to Compress Evidence for Long-Context UnderstandingPreprint
Xia, Z., Xie, Q., Hue, Y. X., Jiang, J., Lu, G., Jiao, Z., Yuan, X., Dai, D., Mo, T., & Zeng, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.31382
Memory as Middleware for Self-Improving AI AgentsPreprint
Jayaram, K. R., Isahagian, V., Muthusamy, V., Thomas, G., Oum, P., Fang, G., & Aravindan, A. V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32091
“Others pro-pose memory representations : OS-style paging of context [37], test-time neural long-term memory [6], episodic memory control [14], hierarchical and agentic memory graphs [43, 48, 51], gist compression of very long contexts [25], and typed memory representations [1, 22, 39].”
CUE-Mem: Benchmarking Long-Term User Memory via Implicit Cues in Multimodal ConversationsPreprint
Hu, Y., Zhao, Y., Long, Z., Fu, X., Ji, M., Zhao, W., Hou, Y., Wang, Q., & Tu, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32574
Positions Are Not Facts: The Mismatch Between KV Caches and MemoryPreprint
Zhou, C., Zhou, Y., Zhang, S., Gao, J., Li, Z., Wu, H., Yu, H., & Wang, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33759
RoutePrism: Tracing Construction Order Effects in Agent MemoryPreprint
Xu, D., Yang, Z., Wu, J., Zhang, S., Zhu, Z., Li, J., Zhang, J., & Ji, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34160
Stashbird: Episode-Grounded Lifecycle Reconciliation for Conversational Agent MemoryPreprint
Biringa, C. U., Yannul, L., Wang, X., Ayala, M., Yi, N., Moyse, A., Manchanda, N., & Gupta, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34242
“A-Mem [29] evolves linked memory notes, Mne-mis [11] combines similarity retrieval with structured selection, and MemOS [16] and Hindsight [10] manage multiple forms of changing memory. iText2KG [30] incrementally constructs knowledge graphs while resolving duplicate entities and relations.”
Beyond Memory Construction: Rethinking Memory Access for LLM-based Conversational AgentsPreprint
Cai, D., Deng, Y., Wang, Y., Shen, Z., & Ren, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33226
COUNTERMEM: World-Model Verified Counter-Factual Memory for Language AgentsPreprint
Pu, H., Tang, R., & Zhang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.31874
“…strategies across tasks, as in Reflexion (Shinn et al., 2023), ExpeL (Zhao et al., 2024), Generative Agents (Park et al., 2023), Voyager (Wang et al., 2024), MemGPT (Packer et al., 2023), Agent Workflow Memory (Wang et al., 2025), A-Mem (Xu et al., 2025), and ReasoningBank (Ouyang et al., 2026).”
Self-citationAgentsensus: Consensus-Compressed Shared Memory for Multi-Agent Story WorldsPreprint
Pan, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.32297
“Variants change what is stored or how it is maintained rather than who owns it: MemGPT (Packer et al., 2023) pages context like an operating system, MemoryBank (Zhong et al., 2024) decays it on a forgetting curve, Reflexion (Shinn et al., 2023) persists verbal self-feedback, and A-MEM (Xu et al., 2025) links an agent’…”
IS IT ALL ABOUT FEELING? RETHINKING PERSONALIZED LEARNING FOR LASTING KNOWLEDGE
Barbieri, U., Filippone, A., & Piceci, L. (2025) · Giornale Italiano di Educazione alla Salute, Sport e Didattica Inclusiva · doi:10.32043/gsd.v9i1.1475
Self-Designed Evaluators and Warm Memory for Long-Horizon AgentsPreprint
Asgari, S., Kıcıman, E., Nunes, L. D. O., & Chandra, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33717
“…Voyager [67] and Agent Workflow Memory [71] accumulate reusable skills and workflows; generative agents [53], MemGPT [50], MemoryBank [91], A-MEM [74] and Mem0 [13] study how memory is organised, consolidated and retrieved; Buffer of Thoughts [77], Dynamic Cheatsheet [60] and Learn-by-Interact…”
Remember Before You're Asked: MemDream for Self-Probing Memory EvolutionPreprint
Lu, M., Wu, M., Jia, R., Luo, Z., & Zhang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34545
Self-citationGenMem: Generative Symbolic Memory for Self-Evolving HarnessPreprint
Jiang, X., Feng, T., Xu, W., Zhang, Z., Yang, Z., Zhang, W., Zhu, R., Chu, X., Zhao, J., & Wang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34633
Self-citationTRACE: Governing Memory Validity in Evolving Multi-Agent SystemsPreprint
Xiong, W., Zhang, S., Gu, S., Tang, B., Li, Z., Xiong, F., Wen, Y., & Wen, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33517
ContextRender: From Execution Dependencies to Agent ContextPreprint
Kashmira, S., Dantanarayana, J. L., Tang, L., & Mars, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.37743
Thinking Before Thinking: Scaling Agentic Inference Through Meta-ReasoningPreprint
Dahal, P., Bakhtin, A., Cohen, T., Chen, Z., Wu, C.-J., Fergus, R., Yih, S., Synnaeve, G., Salakhutdinov, R., Arora, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.38147
Semantic Projection for Continual Self-Evolution of Language AgentsPreprint
Liu, Z., Chen, J., & Wang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36626
“Recent memory systems organize experience into adaptive notes, linked memories, and retrieved reasoning strategies (Suzgun et al., 2025; Xu et al., 2025; Ouyang et al., 2025; Wei et al., 2025).”
MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy OptimizationPreprint
Wu, J., Yang, Y., Huang, Y., Liu, C., Wang, Q., Qian, C., Wu, S., Zhang, J., & Wang, X. E. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36435
SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use SafetyPreprint
Chen, J., Yu, X., Agrawal, S., & Yu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36201
How Can Recommendation Feedback Evolve Agent Memory?Preprint
Mao, S., Li, M., Zhang, H., Li, Z., Wang, Y., Yu, P., & Zhu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.37544
UpliftMem: Learning Set-Level Uplift for Agent Memory RetrievalPreprint
Liang, M., Qiang, H., Liu, G., & Wu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36805
Memory Is a Derivation: The Distributed-Evidence Paradox in Long-Term AgentsPreprint
Liu, H., & Zhao, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36130
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.