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 901–1,000 of 1,030
RankEvolve: A Reliable Multi-Agent Auto-Research Harness for Evolving Ranking ModelsPreprint
Chen, Z.-Y., Liu, L.-F., Li, H., & Yan, H. (2026) · arXiv.org · doi:10.48550/arxiv.2609.39551
TAGGRAPH: Tag-Augmented Graphs for Graph Retrieval of Agent Persistent HistoriesPreprint
Chen, Y.-S., Liang, Y.-J., & Xie, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.38353
Action Conditioned Bisimulation For GUI Agent MemoryPreprint
Zhang, H., Song, L., Li, Q., Yang, D., Wen, Y., & Yao, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.38778
Personalized State-Transition-Aware Memory for Clinical Agents
Haghifam, M., Rajabi, Z., Sun, Y., & Morato, C. (2026) · arXiv
“A-MEM organizes memories through indexing, linking, and memory evolution (Xu et al., 2025), while Mem0 extracts and consolidates salient information and supports graph-based conflict handling (Chhikara et al., 2025).”
Behnam, A., Kim, S., Yu, J., Huang, E., & Yang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.01780
Heavy-Tailed Memory Traces in Long-Horizon Language AgentsPreprint
Song, X., & Cai, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.00010
“• We characterize heavy-tailed memory usage in”
Deng, W., & Song, X. (2026) · International Journal of Advanced AI Applications
IncrConsist: Efficient Incremental Consistency Maintenance for LLM-Maintained Knowledge Bases
Zhang, B. (2026) · Computing&AI Connect · doi:10.69709/caic.2026.155251
MemTrace: State-Consistent Memory for Long-Horizon Coding AgentsPreprint
Xu, H., Zhou, L., Jin, Z., Zhang, X., Tang, B., Li, Z., Zhou, X., & Zhang, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.04838 · Cited by 1
Test-Time Adaptation of Reasoning Strategies with Bayesian Nonparametric MemoryPreprint
Ramji, K., Naseem, T., & Astudillo, R. F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.06516
“As these memories grow, their contents must also be managed and revised: MemGPT manages information beyond the context window (Packer et al., 2023), Mem0 consolidates conversational records (Chhikara et al., 2025), and A-MEM links and updates structured notes (Xu et al., 2025).”
CIPO: Counterfactual Imagination Policy Optimization for Adaptive Tool Granularity SelectionPreprint
Li, Y., Wan, Y., Zhu, Z., Luo, H., Ren, C., Li, L.-F., & Feng, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.04991
“…reflections, or task records to guide the current agent (Park et al., 2023, Shinn et al., 2023, Zhao et al., 2024, Packer et al., 2023), but the retrieved experience usually remains in the context and does not change the action space (Xu et al., 2025a, Chhikara et al., 2025, Fang et al., 2025).”
Bajaj, R. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7569459
Alasti, A. (2026) · Institutional Repository of Leibniz Universität Hannover (Leibniz Universität Hannover) · doi:10.15488/22777
HGP: An on-device personalized agent memory via hybrid graph storage
Zhou, R., Han, X., Liu, J., Zhang, Y., Meng, F., Feng, J., & Ren, Y. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.406
Self-citationPERSIST: Who-What-When Memory Across Sessions for Full-Duplex Spoken DialoguePreprint
Lin, A., Hou, S., Yu, W., Zhao, X., Niu, H., Geng, W., Xiao, L., Xiao, S., Zhao, M., & Zhang, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.07725
Evolutionary Multi-Agent Collaboration for Real-World Video Face Restoration
Tang, B., Wang, T., Zhang, M., Yu, X., Chen, J., Li, B., & Zhang, K. (2026) · doi:10.1109/cvprf71909.2026.00785
“Recent advancements have led to the development of interactive and dynamic memory systems [23, 45, 55], which not only enable agents to retrieve information but also allow them to organize, update, and reason over their knowledge.”
MINDSET: Energy-based Memory Evolution for Long Conversational AgentsPreprint
Dutta, S., Tummala, S., Vanga, S., Pavani, A., Subrahmanyam, B., & Choudhary, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.08586
“A-MEM dynamically organizes agent-generated notes (Xu et al., 2025), MemoryOS introduces an operating-system abstraction (Kang et al., 2025), Mem0 emphasizes scalable extraction and consolidation (Chhikara et al., 2025), HippoRAG uses graph-based associative retrieval (Gutiérrez et al., 2025), and…”
When to Remember, When to Abstain: Category-Conditioned Retention for Reliable Agent MemoryPreprint
Owolabi, O., Gupta, P., & Wang, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.07100
“…memory systems concentrate on extracting, consolidating, and retrieving information (Packer et al., 2023; Park et al., 2023; Zhong et al., 2024; Xu et al., 2025), and increasingly structure it by semantic type, whether by classifying entity types (Chhikara et al., 2025) or by mining and…”
SkillCycle: Co-Evolving Agent Policies and Skill BanksPreprint
Li, L., Shen, Q., Jiang, Z., Ma, Q., Liu, Y., & Deng, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.09430
“A-Mem dynamically organizes memories through indexing and linking (Xu et al., 2025); Agentic Context Engineering maintains evolving playbooks through generation, reflection, and incremental curation (Zhang et al., 2025).”
CANDO: Cooperative Agentic Network for Layout Design OptimizationPreprint
Masouris, A., Jing, Z., Chandler, B. S., & Jamali‐Rad, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.10044
“Frameworks, such as Generative Agents [37], MemGPT [36], and A-Mem [55], further showed that episodic memory management is essential for coherent long-horizon behavior.”
Use and Disuse: Intent-Structured Experience Consolidation for Memory and Learning in LLM AgentsPreprint
Zeng, X., Liu, B., Wang, X., Jin, Z., Li, Y., & Liu, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.12124
“On LoCoMo, Hippocam achieves 77.6% overall accuracy, exceeding the strongest baseline, A-Mem, by 10.1 pp (Table 3).”
InfluentialEvent-Centric Memory with Query-Aware Graph Augmentation for Long-Term Conversational AgentsPreprint
Liu, Y., Yuan, C., Liu, H., Li, W., Lin, Z., Li, B., Chen, X., & Hu, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.11920
“A-MEM organizes memory notes into an evolving network through dynamically generated links and updates [25].”
DPPM: Dual-Path Parametric Memory for Personalized Language ModelsPreprint
Chen, Y., Liu, S., Shi, J., Hong, J., Cheng, C., Ding, X., Wang, T., Li, Y., Liu, Q., & Xu, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.11776 · Cited by 1
“Mem0 and A-MEM address these needs through memory extraction, organization, and updating (Chhikara et al., 2025; Xu et al., 2025).”
Gated Memory: Admission-Controlled Memory Formation for Conversational AIPreprint
Saraswat, P., Neelagiri, D., & Manoj, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.11270
Clarify, Then Focus: Statement Normalization for Conversation Analytics at ScalePreprint
Arbuzov, M. L., Dave, K., Dontsova, E., Hu, Y., Lao, V., Jain, N., Bei, S., & Dimov, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.10758
“It is evaluated on LoCoMo, whose conversations average about 9,000 tokens across as many as 35 sessions (Xu et al., 2025).”
Agentic Large Language ModelsPreprint
van Duijn, M. J., van der Meer, M., Plaat, A., & van Stein, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.13455
VETR: Execution-Grounded Self-Improving Agent for Symbolic Graphics ReasoningPreprint
Ma, Y., Ai, K., Dai, G., Yu, C., Wang, C., & Shen, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.13098
“…Self-Debugging, and CodeAct (Chen et al., 2024; Wang et al., 2024); VDebugger leverages step-wise traces (Wu et al., 2024); visual unit tests guide revision (Panagopoulou et al., 2025); and agent-memory research examines experience reuse (Hu et al., 2026; Xu et al., 2025; Zhang et al., 2026).”
Liu, S., Li, H., & Qian, J. (2026) · Electronics · doi:10.3390/electronics15204688
Forgetting is Not Always Bad: A Neuro-Inspired Memory Repair Mechanism for Poisoned LLM AgentsPreprint
Liu, L., Liang, Y., Zhang, X., Liu, J., Li, Q., Guo, B., & Yu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.13267
“Inspired by the Ebbinghaus Forgetting Curve theory [42], we model this process as exponential decay in Eq.”
ARTEMIS enables reusable analytical context generation from scientific literature
Li, D., Gong, B., Tong, W., & Xu, J. (2026) · Discover Artificial Intelligence · doi:10.1007/s44163-026-02485-5
“Prior work has highlighted the value of retrieval [12–14], grounding [7, 15], memory, and iterative synthesis for improving LLM-based reasoning over external knowledge [5, 16, 17].”
VitaBench: Memory Lifecycles Beyond RetrievalPreprint
Yu, M., & Hu, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.14731
Li, R., Chen, F., & Li, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.14795
“…the core LLM typically handles high-level reasoning, planning, and decision-making [1], while auxiliary modules, such as external tools [3, 22] and memory or experience components [23], allow agents to interact with real or simulated environments [24] and accumulate long-term experience [10].”
Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon AgentsPreprint
Lu, Y., McAuley, J., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.18784
Seeing, Maintaining, and Learning: The Evolution of Multimodal Agent MemoryPreprint
Guo, M., Jiao, Q., Liu, X., Pang, J., Yang, R., Xu, N., Che, L., Xue, H., Wang, Z., Cai, X., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.19366 · Cited by 2
Self-citationAlphaDiana: A Measurement System for Agents in Verifiable Reasoning TasksPreprint
Zhou, Z., Li, Z., Huang, W., Li, X., Cao, C., Feng, X., Lu, X., Hu, J., Xie, Y., Pelleriti, N., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.20675 · Cited by 1
TEM: Temporal State Tracking Memory for Large Language Model AgentsPreprint
Song, P., Zhuang, Y., & Shi, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.21900
“A-MEM, Mem0 and MemoryOS extend this work through memory associations, updates and hierarchical organisation [10–12].”
Memorilla: Latent Semantic Memory for LLMsPreprint
Fallahpour, A., Idehpour, P., Kothapalli, V., Mounier, N., Wu, S., Pardis, S., & Leskovec, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.21375
“MemGPT (Packer et al., 2023) moves information between context tiers, MemoryBank (Zhong et al., 2024) stores conversations and summaries, and Mem0 (Chhikara et al., 2025) and A-MEM (Xu et al., 2025) maintain structured records across sessions.”
Librarian Agents: Training LLMs to Curate Long-Context Workspaces that Anticipate Future QueriesPreprint
Singh, A., Zhang, T., Lee, Y., Nair, R. S., Kumar, A., & Finn, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.22273
Zou, W., Yi, P., & Luo, S. (2026) · International Conference on Intelligent Human-Machine Systems and Cybernetics · doi:10.1109/ihmsc70922.2026.00027
A Hierarchical Memory-Based Exploration Mechanism for Creative Language Model Agents
He, C., Bi, Z., Lou, J., Hu, Z., Xue, Z., Shen, Q., Liu, Z., & Zhao, K. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11729301
“In knowledge representation, research has progressed from simple vector storage to complex structural networks; A-MEM [24] constructs interconnected knowledge graphs, while Mem0 [25] employs graph-based representations to capture entity relationships.”
HMem: A Hawkes Process Memory Framework for Long-Horizon Agents
Yan, Z., Wu, X., & Wei, J. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11728985
From Relevant Memories to Sufficient Evidence: Rethinking Memory Utilization in LLM AgentsPreprint
Li, S., Yang, W., Zhang, P., Cheng, A., Ping, H., Zhang, X., Thomason, J., Ma, X., Zhao, Y., & Bogdan, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.22668
“Mem0 (Chhikara et al., 2025) extracts persistent facts from dialogue, while A-Mem (Xu et al., 2025) develops interconnected memory notes that evolve as new experiences arrive.”
InfluentialMARS: Memory-Aware Routing for Session-Level LLM AssistantsPreprint
Wang, M., Zuo, J., Dai, Z., Wang, Z., & Kong, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.23583
“We use M EM 0 and A-MEM as representative plug-in memory substrates because they can be consistently crossed with fixed generators and routers across both benchmarks.”
InfluentialAgentic Memory Tagging: Scalable Yet Stable Indexing for Long-Term Memory
Zhu, S., Fan, M., Zhou, W., Ma, C., Liu, X., & Xiong, H. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11728438
HIM: A Human-Inspired Memory Loop for LLM Agents via Encoding, Consolidation, and Retrieval
Tang, H., Ye, L., & Xie, S. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11728937
“In practice, the core difficulty is not simply “storing more,” but using memory reliably: the agent must preserve attribution (who said what), resist interference from outdated traces, and avoid retrieval noise as memory grows [7,8].”
Hippo: Training-Free Agent Self-Evolution via Hippocampal ReplayPreprint
Song, Y., Wu, J., Ren, Y., Qin, T., & Wang, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24453
“Mem0 extracts and updates salient conversational information (Chhikara et al., 2025), while A-MEM updates the descriptions and attributes of related memories (Xu et al., 2025).”
Is This Your Agent? Identity Watermarking for Agent Systems in Dynamic EnvironmentsPreprint
Zhong, Y., Cheng, Y., Zhou, J., Liu, W., Xie, Y., Hu, Y., Xie, Y., & Yin, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24382
“Each domain is evaluated under four runtime backends: No-Mem operates without memory, A-MEM (Xu et al., 2025) continuously updates memory, Memento (Zhou et al., 2025) evolves retrievable historical cases, and Evolving Harness (Cheng et al., 2026a) derives and updates prompting rules and skills from…”
InfluentialMemDerive: A Diagnostic Benchmark for Agent Memory over Evolving DataPreprint
Gao, S., Li, Y., Chen, H., He, B., Wu, M., Zeng, J., Heinrich, M., & Sun, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.25373
MemPrice: Online Memory Allocation via Adaptive Pricing for Long-Horizon AgentsPreprint
Zong, Y., Piao, S., Jiang, J., & Tsung, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24955
“(Chhikara et al., 2025); and A-MEM organizes memories into an evolving network whose representations and associations can change as new experiences arrive (Xu et al., 2025).”
Memories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM AgentsPreprint
Wang, Y., ju, J., Zhang, J., Wu, Z., Duan, Y., Sun, Z., Cai, T., & Zheng, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.25123
“Agentic memorieslink notes by semantic relatedness [24].”
DuplexLedger: Benchmarking Transactional Memory in Full-Duplex Speech AgentsPreprint
Li, Y., Guan, X., Li, K., Wang, J., Luo, K., Xiao, Y., Guo, Y., Chen, Z., Lin, L., & Wang, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.27581
“…instead study long-horizon storage, retrieval, and updating from serialized histories (Packer et al., 2023; Maharana et al., 2024; Wu et al., 2025; Xu et al., 2025), while recent speech agents begin to connect persistent memory with streaming interaction (Yao et al., 2025; Xie et al., 2026).”
MemHRL: Aligning Hierarchical Memory and Policy Learning for Long-Horizon LLM AgentsPreprint
Yin, X., Huang, C., Liu, R., & Ning, Y. (2026) · Research Square · doi:10.21203/rs.3.rs-11435233/v1
What should AI forget? A design and taxonomy for negative space governance in AI memoryPreprint
Xu, E., Ding, K., Chang, T.-Y., Li, T., & Bashir, M. N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.27607
Santos, G. M., Julia, R. M. S., & Nascimento, M. (2026) · Simpósio Brasileiro de Jogos e Entretenimento Digital (SBGames) · doi:10.5753/sbgames.2026.25637
“MemGPT [Packer et al. 2023] treats the LLM context as a virtual-memory page cache backed by external storage; A-MEM [Xu et al. 2025] structures memory as a Zettelkasten of dynamically linked notes; Reflexion [Shinn et al. 2023] casts verbal self-reflection as an episodic buffer; and Voyager [Wang…”
InfluentialCapability-Aware Dynamic Skill Selection for Memory-Augmented Generation
Qiao, Y., Wang, Y., Huang, C., Zhang, X., & Shang, S. (2026) · International Conference on Information Science and Technology · doi:10.1109/icist71072.2026.11727740
“Other systems such as A-MEM [20] and MemoryOS [21] explore self-evolving memory structures and OS-style memory abstractions.”
Coherent Generative Agentic AI Enabled 6G/7G Networking
Galis, A. (2026) · Future Internet · doi:10.3390/fi18100576
SpatialBrain: Spatial Intelligence Growing from ExperiencePreprint
Zhou, Y., Gong, Z., Ye, X., Liu, Y., Li, D., Zhou, X., Zhong, Z., & Yang, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.29317
Remembering Through Change: Provenance-Preserving Hypergraph Memory for Multimodal AgentsPreprint
Tang, J., Zeng, L.-A., Yuan, X., Dong, X., Xia, Z., Li, Q., Kai, S., Liang, K., Yao, R., & Yin, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.29995
“We compare with four unimodal memory methods using dense image caption: A-Mem (Xu et al., 2025), MemoryOS (Kang et al., 2025), SimpleMem (Liu et al., 2026c), and HyperMem (Yue et al., 2026); and four multimodal memory or RAG methods: M2A (Feng et al., 2026), MMA (Lu et al., 2026), SRAG (Guo et al.,…”
InfluentialTsang, H. T., Zhang, B., Huang, H., Bai, J., Li, Y., Gao, Y., Xie, Z., Luo, H., Pan, J. Z., & Song, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.29249
“A-MEM caption text is the closest to balanced, but no system occupies the upper-right.”
InfluentialTransparent, Traceable, Deterministic: Agentic Memory via Knowledge Graphs
Gentile, A. L., An, S., & DeLuca, C. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-42029-9_16
ColaBench: Evaluating Memory-Guided Knowledge Acquisition in Personal AgentsPreprint
Zhang, W., Shi, Z., Jiang, X., Wu, X., Yang, C., Xu, C., & Song, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.30989
“A-MEM builds evolving links between memory notes [17], while Graphiti, the graph engine underlying Zep, maintains temporal entity and fact relations [12].”
CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPUPreprint
Pan, D., Gao, X.-J., Fan, J.-Y., Du, B., Li, J.-S., Chen, X.-X., Chang, N.-Y., Li, X., Zheng, L.-C., Li, J., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.31034
“Agent memory has progressed from per-episode self-reflection [18] and single-agent skill or experience libraries [19, 20] to self-linking memory notes [21], and recently to sharing across actors through banked lessons for cooperating code agents [22] and provenance-tracked multi-user memory [23].”
The Value of Memory in Language AgentsPreprint
Lin, Z., Zhu, E., Hu, H., Hao, X., Li, C., Liu, L., Chen, K., Du, C., & Xiong, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.31303
“…Agents connects recorded experience and reflection to planning (Park et al., 2023); MemGPT manages transfers across memory tiers (Packer et al., 2023); A-Mem evolves linked notes (Xu et al., 2025); and Mem0 extracts, consolidates and retrieves conversational memories (Chhikara et al., 2025).”
Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress TestPreprint
Liu, B., Wang, Y., Shen, B., & Shang, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.30698
“MemGPT introduced virtual-context management for persistent agents (Packer et al., 2023); Mem0, A-Mem, and LangMem develop extraction, organization, consolidation, and retrieval mechanisms (Chhikara et al., 2025; Xu et al., 2025; LangChain, 2025).”
Babu, M., Nejatia, F., & S, M. E. (2026) · Research Square · doi:10.21203/rs.3.rs-11226881/v1
Reinforcement Learning for the Working Memory Controller
Gokulraj, J., & Karthikeyan, L. (2026) · International Conference on Smart Electronics and Communication (ICOSEC) · doi:10.1109/icosec70158.2026.11735862
SPECMEM: Accelerating Agentic Tool Calling via Live Memory ManagementPreprint
Song, Z., Zhao, W., Liu, Y., Zhu, Y., Xu, K., Li, S., & Yu, P. S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.31840
“Agent memory systems keep bounded, updated, user-scoped stores (MemGPT (Packer et al., 2023), MemoryBank (Zhong et al., 2023), Generative Agents (Park et al., 2023), Reflexion (Shinn et al., 2023), A-MEM (Xu et al., 2025), Agent Workflow Memory (Wang et al., 2024)), but for knowledge or skill recall, not speculative d…”
ActionRAG: Evidence-Refining Graph Diffusion for Long-Term Memory in LLM AgentsPreprint
Mahla, N., & Vandrangi, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.32752
“A-MEM (Xu et al., 2025) adds agentic links and a self-organising memory graph and Zep (Rasmussen et al., 2025) maintains a temporal knowledge graph for evolving facts.”
Shimomoto, T. (2026) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.20379919
A controlled embedder swap on LoCoMo, and three arms that could not carry a comparison
Bering, A. (2026) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.22714834
Seamless: Self-Maintaining Knowledge Graph Memory for Retrieval-Augmented Language Models
Smirnov, N. (2026) · IEEE Access · doi:10.1109/access.2026.3747787
Selective Rigidity: An Impossibility Result and Benchmark for Identity-Preserving Agent Learning
Chakraborty, S., Srivastava, A., & Nandu, S. (2026) · IEEE International Conference on Artificial Intelligence, Computer Science and Information Processing (AICSIP) · doi:10.1109/aicsip69702.2026.11744034
“CoALA [1] and MemGPT [2] organise agent memory into modular stores; subsequent systems (A-MEM [3], HippoRAG [4], GraphRAG [15], Self-RAG [16], Larimar [17]) add richer organisation and retrieval.”
Adaptive Memory Index Evolution for AgentsPreprint
Hu, Y., Su, T., Sun, Y., Ma, S., Tan, Q., & Zhao, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34977 · Cited by 1
FlexA-Mem: Budget-Adaptive Agentic Memory for LLM AgentsPreprint
Xu, W., Han, J., Zhu, X., Zhang, Q., Liang, K., Yin, Y., Mei, K., & Metaxas, D. N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.33274
“A-Mem (Xu et al., 2025a) fills the allowance with retrieved notes in raw text, and Summary+A-Mem adds a running LLM summary (Wang et al., 2025c) that takes half of the allowance on LongMemEval and 70% on the agentic tasks.”
InfluentialSelf-citationConMem: Counterfactual Multi-Scenario Probing for Resolving LLM Memory MisalignmentPreprint
Zhang, S., Ji, T., Chen, J., Xu, E., Chu, B., & Yi, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.33432 · Cited by 1
“• A-MEM (Xu et al., 2025): organizes dialogue into attributed memories.”
InfluentialBehaviorMem: Building and Evaluating User Memory from Real-World E-Commerce Behavior StreamsPreprint
Wang, S., Fang, X., Qiu, R., Ma, Z., Chen, D., Kuang, K., & Guo, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34790
StreamVibe: Proactive Video Agents for Evolving User IntentPreprint
Liu, C., Lai, Y., Yang, Y., Liu, Y., Hou, J., Cao, Y., Wang, H., Fan, R., Wang, L., Liu, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34662
Agent Memory as Recommendation: Recommender-System Design for Long-Term Memory of LLM AgentsPreprint
Zhang, X., & Sun, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34673
“Memory systems build different structures, such as linked notes, knowledge graphs, temporal hierarchies and fact-augmented keys [17, 23, 40, 42].”
CaliMem: Search-Efficient Data Agent via Calibrated MemoryPreprint
Li, H., Li, A. Z., Kang, F., Wang, J. T., Chen, Y., Liu, R., & Jia, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.34500
“…preserve experience as verbal reflections (Shinn et al., 2023), distilled workflows (Zhao et al., 2024; Wang et al., 2025b), reusable skills (Wang et al., 2024), evolving playbooks (Suzgun et al., 2026; Zhang et al., 2025b), or structured reasoning memories (Xu et al., 2025; Ouyang et al., 2026).”
EMMA: Structured Episodic Memory for Multi-Agent Workflow SynthesisPreprint
Niu, R., Liu, K., Yu, R., & Klinkner, K. L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.35966
“Reflexion, A-Mem and MemRL store retrievable reflections, notes and utility-ranked entries [Shinn et al., 2023, Xu et al., 2025, Zhang et al., 2026a], while ACE and GEPA adapt the context of a ReAct-style agent, as an accumulating playbook and as a reflectively optimized prompt [Zhang et al.,…”
ETAMem: Decoupling and Aligning Heterogeneous Memory for Long-Term Conversational AgentsPreprint
Qian, X., Zhang, Y., Zhao, Y., Liu, X., Zhou, B., & Pan, J. Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.37443
Self-citationCausal Decomposition of LLM Agent Failure via Oracle-State Intervention
Jia, L., Saadat, M., Wong, M., Venkatesh, A., & Nijjer, K. (2026) · Workshop for Research on Agent Language Models (REALM) · doi:10.18653/2026.realm-1.6
ExGQL: Self-Improving Natural Language-to-Graph Query Generation via Experience-Guided Learning
Guo, P., Du, B., Zhou, H., Gupta, C., Gu, Y., Küllian, R., & Jiang, S. (2026) · Workshop for Research on Agent Language Models (REALM) · doi:10.18653/2026.realm-1.58
MemPulse: Memory Pulse-Driven Adaptive Long-Context Reasoning with Fact-based Hypergraph
Tao, W., Wang, Y., Lan, Y., & Yang, C. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.947
GRACE-Mem: Graph Retrieval with Agentic Curation of Evidence for Long-Term Conversational Memory
Wang, H., Chen, X. N., Chang, P.-Y., Chen, C.-H., Huang, H.-Y., Lin, C., & Chung, I.-L. (2026) · Automated Knowledge Base Construction · doi:10.18653/2026.akbc-1.21
Liu, J., Liu, T., & Zhang, C. (2026) · Reliability Engineering & System Safety · doi:10.1016/j.ress.2026.113770
When Do Memories Break? Chain-of-Stage Diagnosis for LLM Memory Systems
Wang, C., Fu, Y., Zheng, Z., Qin, S., Cao, Z., & Liu, Z. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.1042
When Should LLM Agents Forget? Memory Filtering Under Query-Critical Resolver Errors
Vinayaka, S. (2026) · Workshop for Research on Agent Language Models (REALM) · doi:10.18653/2026.realm-1.16
ATC: Agent Trajectory Control for Role-Contract Enforcement in Black-Box Multi-Agent Systems
Maliha, M., Pramanik, V., & Jha, S. K. (2026) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/2026.emnlp-main.2058
GraP-Mem: Granularity-Aware Planning for Adaptive Memory Access in Long-Horizon LLM Agents
Zeng, Q., Dong, J., Lu, W., & Du, X. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.2375
Prism: Persistent Faceted File Memory for Cost-Efficient Coding Agents
Zhang, Z., Liu, Y., Zhang, Y., & Huaqian, C. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.1301
Self-citationDo Query-Agnostic Memory Structures Earn Their Budget? A 10M-Token Audit with Exact Read Limits
Maraj, A., & Sethi, K. (2026) · Workshop for Research on Agent Language Models (REALM) · doi:10.18653/2026.realm-1.18
EffectMem: Effectiveness-Driven Memory Reuse for Long-Horizon LLM Agents
Yin, J., Wang, L., Cao, D., & Hong, Y. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.859
Capture Is Easy, Forgetting Is Hard: Benchmarking Prospective Memory in Conversational Agents
Bilgin, D., Schewior, J., & Wendemuth, A. (2026) · Workshop for Research on Agent Language Models (REALM) · doi:10.18653/2026.realm-1.43 · Cited by 1
Cross-Domain Transfer of the Procedural Skill Cache
Gokulraj, J., & Karthikeyan, L. (2026) · International Conference on Innovative Mechanisms for Industry Applications · doi:10.1109/icimia69818.2026.11744954
“…structural abstraction, whether transfer utility is predicted by a learned estimator rather than a similarity proxy, whether a A-MEM [21] No No No No CDST (ours) Yes Yes Yes Yes As Table I shows, Voyager caches skills without abstraction or adaptation, Dream Coder builds structural…”
Is it my preference? Role Grounding Attack against Agent's Memory via Social InteractionPreprint
Bu, H., Shen, Y., Zhang, L., Hu, R., & Zhang, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.38209
MemSteer: Turning User Histories into Supervision for Agent Memory OptimizationPreprint
Huang, J., Lan, C., Yu, Q., Peng, X., & Lu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.39276
EmbodiExp: Compositional Experience Transfer for Embodied AgentsPreprint
Gu, Z., Chen, Y., Jiang, R., Yi, L., & Wang, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.39180
“…as episodic records, summaries, and structured knowledge (Park et al., 2023; Packer et al., 2023; Zhong et al., 2024; Chhikara et al., 2025; Xu et al., 2025; Zhang et al., 2025), with extensions for multimodal consolidation (Liu et al., 2025) and learnable memory operations (Zhang et al.,…”
CALM: Conflict-Aware Lifecycle Memory Governance for Long-Horizon Language AgentsPreprint
Li, C., Ou, X., Ma, X., Li, H., Xu, H., Liang, J., Wang, X., Chai, J., Yin, G., & Liu, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.38573
PyRAMIND: Auditable Deterministic Pyramid Memory for Long-Running LLM Agents
Roy, A., Upadhyay, R., Millar, A. P., Murray, L., & Mills, R. (2026) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3799682.3840047
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.