pith. sign in

arxiv: 2606.19847 · v1 · pith:EZECV7GZnew · submitted 2026-06-18 · 💻 cs.CL

AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts

Yanyu Yao , Shangze Li , Zhi Zheng , Hui Zheng , Qi Liu , Tong Xu , Enhong Chen This is my paper

Pith reviewed 2026-06-26 17:49 UTC · model grok-4.3

classification 💻 cs.CL
keywords LLM agentslong-term memoryatomic factsmemory systemsLoCoMo benchmarkFact Executor
0
0 comments X

The pith

AtomMem extracts atomic facts to create stable long-term memory for LLM agents.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Current LLM memory systems build coarse and unstable representations that hinder long-term information reuse across sessions. AtomMem addresses this by introducing a Fact Executor to pull out high-value atomic facts from interactions. These facts are then organized into hierarchical event structures and temporal profiles, with an associative graph for retrieval. This design aims to deliver value-dense storage and stable evolution of memory. Experiments show it reaches state-of-the-art results on the LoCoMo benchmark for various reasoning tasks.

Core claim

AtomMem introduces a Fact Executor that selectively extracts high-value atomic facts from long-form interactions to serve as efficient memory representations. It organizes these facts into hierarchical event structures and temporal profiles for coherent contexts and evolving user attributes, activating an associative memory graph during retrieval.

What carries the argument

The Fact Executor, which selectively extracts high-value atomic facts from long-form interactions to enable value-dense and stable memory storage.

If this is right

  • Allows LLM agents to accumulate and reuse information over multiple sessions without context window limits.
  • Organizes memories hierarchically to capture episodic contexts and track user attributes over time.
  • Uses an associative memory graph to connect fragmented memories during retrieval.
  • Achieves state-of-the-art performance on reasoning tasks in the LoCoMo benchmark.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This approach could lower the computational cost of maintaining long-term agent memory compared to full conversation logs.
  • The method might generalize to other sequential data processing tasks beyond LLM agents.
  • Testing on additional benchmarks could reveal if the hierarchical organization provides advantages in specific domains like personal assistance.

Load-bearing premise

The Fact Executor can reliably and selectively extract high-value atomic facts from long-form interactions in a manner that is both value-dense and free of the instability seen in unconstrained memory updates.

What would settle it

A direct comparison where the Fact Executor is replaced with random or full-context storage, showing whether performance on LoCoMo drops significantly.

Figures

Figures reproduced from arXiv: 2606.19847 by Enhong Chen, Hui Zheng, Qi Liu, Shangze Li, Tong Xu, Yanyu Yao, Zhi Zheng.

Figure 1
Figure 1. Figure 1: Architecture comparison. AtomMem overcomes the bloated storage and isolated matching of previous methods by organizing atomic facts into associative graphs for precise hierarchical retrieval. rely on frequent LLM-driven rewrites to update existing entries. While this design enables flexible knowledge organization and continuous adaptation, unconstrained updates introduce severe instability. Hallucinations … view at source ↗
Figure 2
Figure 2. Figure 2: The overall architecture of AtomMem. It is designed to support high-density memory storage, stable user-state evolution, and efficient retrieval for long-term personalized agents. before they enter the memory system, while en￾suring that each generated fact is independent and comprehensible without external context. 3.1.2 Structured Fact Construction While the Atomic Fact Extractor provides clean tex￾tual … view at source ↗
Figure 3
Figure 3. Figure 3: Performance sensitivity analysis under vary [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: System prompt for atomic fact extraction. The prompt instructs the fact executor to filter low-value content, resolve references, rewrite extracted information as standalone third-person facts, integrate multimodal evidence, and output the extracted facts in JSON format. temporal proximity. A.3.1 Query-Aware Keyword Weighting As defined in Section section 3.4, the entity edge weight relies on a query-aware… view at source ↗
Figure 5
Figure 5. Figure 5: Training sample from dataset D. The example shows an instruction-output pair used for training the fact executor to extract standalone, high-value third-person facts from dialogue. Algorithm 1 Event Memory Construction and Update 1: Input: Verified new fact Fnew, retrieved context facts Cret with top-k facts 2: Output: Updated memory system state 3: Initialize event candidate set Ecand ← ∅ 4: Initialize st… view at source ↗
Figure 6
Figure 6. Figure 6: Hyperparameter analysis of the graph retrieval [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Hyperparameter analysis of the compensatory [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: System Prompt for Response Generation. This prompt is utilized for single-hop, multi-hop, and temporal reasoning tasks where precise extraction is required. Figure A.4: System Prompt for Response Generation - Open Domain Task: Generate an answer based on retrieved information (Profiles and/or Facts). Input: - query: Original user query - profiles: List of Profile statements (optional) - facts: List of Fact… view at source ↗
Figure 9
Figure 9. Figure 9: System Prompt for Response Generation (Open Domain). This prompt guides the model to integrate retrieved memory with external knowledge for comprehensive reasoning. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: System Prompt for Answer Judgment. This prompt configures the LLM judge to evaluate generated answers against ground-truth references. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
read the original abstract

Large language models (LLMs) demonstrate strong reasoning and generation abilities, but their fixed context windows limit long-term information accumulation and reuse across multi-session interactions. Existing memory-augmented systems often construct memory in a coarse and unstable manner, relying on inefficient memory representations or unstable unconstrained updates. To address these challenges, we propose AtomMem, a long-term memory system designed for value-dense storage and stable memory evolution. AtomMem introduces a Fact Executor, which selectively extracts high value atomic facts from long form interactions to serve as highly efficient memory representations. Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time. During retrieval, the system activates an associative memory graph to connect fragmented memories. Experiments on the LoCoMo benchmark confirm that AtomMem achieves state-of-the-art performance across various reasoning tasks, offering a scalable and economically viable solution for deploying intelligent personalized agents.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes AtomMem, a long-term memory system for LLM agents. It introduces a Fact Executor to selectively extract high-value atomic facts from long-form interactions as efficient memory representations, organizes these into hierarchical event structures and temporal profiles for episodic context and user attribute tracking, and employs an associative memory graph for retrieval. The central claim is that experiments on the LoCoMo benchmark demonstrate state-of-the-art performance across reasoning tasks, providing a scalable solution for personalized agents.

Significance. If the SOTA results on LoCoMo are substantiated with proper controls, AtomMem could offer a practical advance in stable, value-dense memory for multi-session LLM agents, addressing fixed context limits. The atomic-fact approach and hierarchical organization represent a targeted design choice worth evaluating against existing memory-augmented systems.

major comments (2)
  1. [Abstract / Experiments] Abstract and Experiments section: The claim that AtomMem achieves state-of-the-art performance on the LoCoMo benchmark supplies no baselines, metrics, error bars, ablation studies, or method details, so the central experimental result cannot be evaluated from the manuscript.
  2. [§3] §3 (Fact Executor description): The Fact Executor is asserted to reliably and selectively extract high-value atomic facts in a manner free of instability seen in unconstrained updates, but the manuscript provides no mechanism details, selection criteria, or empirical validation of this stability property, which is load-bearing for both the design and the SOTA claim.
minor comments (1)
  1. [Abstract] Abstract: 'long form interactions' and 'high value atomic facts' should be hyphenated for consistency with technical writing.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments point by point below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: [Abstract / Experiments] Abstract and Experiments section: The claim that AtomMem achieves state-of-the-art performance on the LoCoMo benchmark supplies no baselines, metrics, error bars, ablation studies, or method details, so the central experimental result cannot be evaluated from the manuscript.

    Authors: We agree that the current manuscript does not supply baselines, metrics, error bars, ablation studies, or sufficient method details to allow evaluation of the SOTA claim on LoCoMo. This is a substantive gap. In revision we will expand the Experiments section with full baseline comparisons, all metrics reported with error bars, ablation studies, and complete method descriptions; the abstract will be updated to summarize these additions accurately. revision: yes

  2. Referee: [§3] §3 (Fact Executor description): The Fact Executor is asserted to reliably and selectively extract high-value atomic facts in a manner free of instability seen in unconstrained updates, but the manuscript provides no mechanism details, selection criteria, or empirical validation of this stability property, which is load-bearing for both the design and the SOTA claim.

    Authors: We acknowledge that §3 currently offers only a high-level description and lacks explicit mechanism details, selection criteria, and empirical validation of stability. These elements are indeed central. We will revise §3 to include a precise account of the Fact Executor’s operation, the criteria used to identify high-value atomic facts, and supporting empirical analyses demonstrating improved stability relative to unconstrained updates. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes an engineering system (AtomMem with Fact Executor, hierarchical structures, and associative graph) validated by benchmark experiments on LoCoMo. No equations, derivations, fitted parameters, or first-principles claims appear in the provided text. Central performance claims rest on external empirical evaluation rather than any self-referential reduction or self-citation chain. This is the expected outcome for a non-derivational systems paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no technical sections available to identify free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5703 in / 1129 out tokens · 23535 ms · 2026-06-26T17:49:51.083929+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 22 canonical work pages · 14 internal anchors

  1. [1]

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2024. https://openreview.net/forum?id=hSyW5go0v8 Self- RAG : Learning to retrieve, generate, and critique through self-reflection . In The Twelfth International Conference on Learning Representations

    2024

  2. [3]

    Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. 2025. https://doi.org/10.48550/arXiv.2504.19413 Mem0: Building production-ready AI agents with scalable long-term memory . Preprint, arXiv:2504.19413

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.19413 2025

  3. [4]

    Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang. 2026. https://openreview.net/forum?id=dyJ0GWpjJB LightMem : Lightweight and efficient memory-augmented generation . In The Fourteenth International Conference on Learning Representations

    2026

  4. [5]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. 2023. https://doi.org/10.48550/arXiv.2312.10997 Retrieval-augmented generation for large language models: A survey . Preprint, arXiv:2312.10997

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2312.10997 2023

  5. [6]

    Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang. 2020. https://proceedings.mlr.press/v119/guu20a.html Retrieval augmented language model pre-training . In Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 3929--3938. PMLR

    2020

  6. [9]

    u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K \"u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \"a schel, Sebastian Riedel, and Douwe Kiela. 2020. https://proceedings.neurips.cc/paper/2020/hash/6b493230205f780e1bc26945df7481e5-Abstract.html Retrieval-augmented generation for knowledge-intensive NLP ...

    2020

  7. [10]

    Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang. 2023. https://doi.org/10.48550/arXiv.2311.08719 Think-in-memory: Recalling and post-thinking enable LLM s with long-term memory . Preprint, arXiv:2311.08719

    work page doi:10.48550/arxiv.2311.08719 2023

  8. [11]

    and Lin, Kevin and Hewitt, John and Paranjape, Ashwin and Bevilacqua, Michele and Petroni, Fabio and Liang, Percy

    Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2024. https://doi.org/10.1162/tacl_a_00638 Lost in the middle: How language models use long contexts . Transactions of the Association for Computational Linguistics, 12:157--173

    work page doi:10.1162/tacl_a_00638 2024

  9. [12]

    Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong, Enhong Chen, and Tong Xu. 2026. https://doi.org/10.48550/arXiv.2603.23231 PERMA : Benchmarking personalized memory agents via event-driven preference and realistic task environments . Preprint, arXiv:2603.23231

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2603.23231 2026

  10. [14]

    Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Sch \"u tze. 2023. https://doi.org/10.48550/arXiv.2305.14322 RET-LLM : Towards a general read-write memory for large language models . Preprint, arXiv:2305.14322

    work page doi:10.48550/arxiv.2305.14322 2023

  11. [15]

    Rodrigo Nogueira and Kyunghyun Cho. 2019. https://doi.org/10.48550/arXiv.1901.04085 Passage re-ranking with BERT . Preprint, arXiv:1901.04085

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.04085 2019

  12. [17]

    MemGPT: Towards LLMs as Operating Systems

    Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G. Patil, Ion Stoica, and Joseph E. Gonzalez. 2023. https://doi.org/10.48550/arXiv.2310.08560 MemGPT : Towards LLM s as operating systems . Preprint, arXiv:2310.08560

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2310.08560 2023

  13. [21]

    Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. 2025. https://doi.org/10.48550/arXiv.2410.10813 LongMemEval : Benchmarking chat assistants on long-term interactive memory . In The Thirteenth International Conference on Learning Representations

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2410.10813 2025

  14. [22]

    Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, and 1 others. 2023. https://doi.org/10.48550/arXiv.2309.07864 The rise and potential of large language model based agents: A survey . Preprint, arXiv:2309.07864

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2309.07864 2023

  15. [23]

    Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, and Mike Lewis. 2024. https://doi.org/10.48550/arXiv.2309.17453 Efficient streaming language models with attention sinks . In The Twelfth International Conference on Learning Representations

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2309.17453 2024

  16. [24]

    Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. 2025. https://doi.org/10.48550/arXiv.2502.12110 A-MEM : Agentic memory for LLM agents . In Advances in Neural Information Processing Systems

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2502.12110 2025

  17. [25]

    Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. 2023. https://doi.org/10.48550/arXiv.2305.10250 MemoryBank : Enhancing large language models with long-term memory . Preprint, arXiv:2305.10250

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2305.10250 2023

  18. [26]

    Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low, and Paul Pu Liang. 2025. https://doi.org/10.48550/arXiv.2506.15841 MEM1 : Learning to synergize memory and reasoning for efficient long-horizon agents . Preprint, arXiv:2506.15841

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.15841 2025

  19. [27]

    2025 , url =

    Xu, Wujiang and Liang, Zujie and Mei, Kai and Gao, Hang and Tan, Juntao and Zhang, Yongfeng , booktitle =. 2025 , url =. 2502.12110 , archivePrefix =

    Pith/arXiv arXiv 2025

  20. [28]

    Mem0: Building Production-Ready

    Chhikara, Prateek and Khant, Dev and Aryan, Saket and Singh, Taranjeet and Yadav, Deshraj , year =. Mem0: Building Production-Ready. 2504.19413 , archivePrefix =

    Pith/arXiv arXiv

  21. [29]

    Think-in-Memory: Recalling and Post-thinking Enable

    Liu, Lei and Yang, Xiaoyan and Shen, Yue and Hu, Binbin and Zhang, Zhiqiang and Gu, Jinjie and Zhang, Guannan , year =. Think-in-Memory: Recalling and Post-thinking Enable. 2311.08719 , archivePrefix =

    arXiv

  22. [30]

    2305.10250 , archivePrefix =

    Zhong, Wanjun and Guo, Lianghong and Gao, Qiqi and Ye, He and Wang, Yanlin , year =. 2305.10250 , archivePrefix =

    Pith/arXiv arXiv

  23. [31]

    2023 , eprint =

    Modarressi, Ali and Imani, Ayyoob and Fayyaz, Mohsen and Sch. 2023 , eprint =

    2023

  24. [32]

    2306.03901 , archivePrefix =

    Hu, Chenxu and Fu, Jie and Du, Chenzhuang and Luo, Simian and Zhao, Junbo and Zhao, Hang , year =. 2306.03901 , archivePrefix =

    arXiv

  25. [33]

    Kang, Jiazheng and Ji, Mingming and Zhao, Zhe and Bai, Ting , booktitle =. Memory. 2025 , address =. doi:10.18653/v1/2025.emnlp-main.1318 , url =

    work page doi:10.18653/v1/2025.emnlp-main.1318 2025

  26. [34]

    and Stoica, Ion and Gonzalez, Joseph E

    Packer, Charles and Wooders, Sarah and Lin, Kevin and Fang, Vivian and Patil, Shishir G. and Stoica, Ion and Gonzalez, Joseph E. , year =. 2310.08560 , archivePrefix =

    Pith/arXiv arXiv

  27. [35]

    Evaluating Very Long-Term Conversational Memory of LLM Agents

    Maharana, Adyasha and Lee, Dong-Ho and Tulyakov, Sergey and Bansal, Mohit and Barbieri, Francesco and Fang, Yuwei , booktitle =. Evaluating Very Long-Term Conversational Memory of. 2024 , address =. doi:10.18653/v1/2024.acl-long.747 , url =

    work page doi:10.18653/v1/2024.acl-long.747 2024

  28. [36]

    GPT-4 Technical Report

    2023 , eprint =. doi:10.48550/arXiv.2303.08774 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2303.08774 2023

  29. [37]

    Sparks of Artificial General Intelligence: Early experiments with GPT-4

    Bubeck, S. Sparks of Artificial General Intelligence: Early Experiments with. 2023 , eprint =. doi:10.48550/arXiv.2303.12712 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2303.12712 2023

  30. [38]

    Llama 2: Open Foundation and Fine-Tuned Chat Models

    Touvron, Hugo and Martin, Louis and Stone, Kevin and Albert, Peter and Almahairi, Amjad and Babaei, Yasmine and Bashlykov, Nikolay and Batra, Soumya and Bhargava, Prajjwal and Bhosale, Shruti and others , year =. doi:10.48550/arXiv.2307.09288 , url =. 2307.09288 , archivePrefix =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2307.09288

  31. [39]

    Transactions of the Association for Computational Linguistics , volume =

    Lost in the Middle: How Language Models Use Long Contexts , author =. Transactions of the Association for Computational Linguistics , volume =. 2024 , doi =

    2024

  32. [40]

    The Twelfth International Conference on Learning Representations , year =

    Efficient Streaming Language Models with Attention Sinks , author =. The Twelfth International Conference on Learning Representations , year =. 2309.17453 , archivePrefix =

    Pith/arXiv arXiv

  33. [41]

    Retrieval-Augmented Generation for Knowledge-Intensive

    Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and K. Retrieval-Augmented Generation for Knowledge-Intensive. Advances in Neural Information Processing Systems , volume =. 2020 , url =

    2020

  34. [42]

    Proceedings of the 37th International Conference on Machine Learning , pages =

    Retrieval Augmented Language Model Pre-Training , author =. Proceedings of the 37th International Conference on Machine Learning , pages =. 2020 , volume =

    2020

  35. [43]

    2023 , eprint =

    Retrieval-Augmented Generation for Large Language Models: A Survey , author =. 2023 , eprint =

    2023

  36. [44]

    Passage Re-ranking with

    Nogueira, Rodrigo and Cho, Kyunghyun , year =. Passage Re-ranking with. 1901.04085 , archivePrefix =

    Pith/arXiv arXiv 1901

  37. [45]

    Active Retrieval Augmented Generation

    Active Retrieval Augmented Generation , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , month = dec, pages =. 2023 , address =. doi:10.18653/v1/2023.emnlp-main.495 , url =

    work page doi:10.18653/v1/2023.emnlp-main.495 2023

  38. [46]

    Asai, Akari and Wu, Zeqiu and Wang, Yizhong and Sil, Avirup and Hajishirzi, Hannaneh , booktitle =. Self-. 2024 , url =

    2024

  39. [47]

    O’Brien, Carrie J

    Generative Agents: Interactive Simulacra of Human Behavior , author =. Proceedings of the 36th Annual. 2023 , publisher =. doi:10.1145/3586183.3606763 , url =

    work page doi:10.1145/3586183.3606763 2023

  40. [48]

    2023 , eprint =

    The Rise and Potential of Large Language Model Based Agents: A Survey , author =. 2023 , eprint =

    2023

  41. [49]

    In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents

    In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents , author =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , month = jul, pages =. 2025 , address =. doi:10.18653/v1/2025.acl-long.413 , url =

    work page doi:10.18653/v1/2025.acl-long.413 2025

  42. [50]

    2025 , eprint =

    Zep: A Temporal Knowledge Graph Architecture for Agent Memory , author =. 2025 , eprint =

    2025

  43. [51]

    Yan, Sikuan and Yang, Xiufeng and Huang, Zuchao and Nie, Ercong and Ding, Zifeng and Li, Zonggen and Ma, Xiaowen and Bi, Jinhe and Kersting, Kristian and Pan, Jeff Z. and Sch. Memory-. 2025 , eprint =

    2025

  44. [52]

    2506.15841 , archivePrefix =

    Zhou, Zijian and Qu, Ao and Wu, Zhaoxuan and Kim, Sunghwan and Prakash, Alok and Rus, Daniela and Zhao, Jinhua and Low, Bryan Kian Hsiang and Liang, Paul Pu , year =. 2506.15841 , archivePrefix =

    Pith/arXiv arXiv

  45. [53]

    2025 , url =

    Wu, Di and Wang, Hongwei and Yu, Wenhao and Zhang, Yuwei and Chang, Kai-Wei and Yu, Dong , booktitle =. 2025 , url =. 2410.10813 , archivePrefix =

    Pith/arXiv arXiv 2025

  46. [54]

    Know Me, Respond to Me: Benchmarking

    Jiang, Bowen and Hao, Zhuoqun and Cho, Young Min and Li, Bryan and Yuan, Yuan and Chen, Sihao and Ungar, Lyle and Taylor, Camillo Jose and Roth, Dan , booktitle =. Know Me, Respond to Me: Benchmarking. 2025 , url =. 2504.14225 , archivePrefix =

    arXiv 2025

  47. [55]

    2603.23231 , archivePrefix =

    Liu, Shuochen and Zhu, Junyi and Shu, Long and Lin, Junda and Chen, Yuhao and Zhang, Haotian and Zhang, Chao and Xu, Derong and Li, Jia and Tang, Bo and Li, Zhiyu and Xiong, Feiyu and Chen, Enhong and Xu, Tong , year =. 2603.23231 , archivePrefix =

    Pith/arXiv arXiv

  48. [56]

    2026 , eprint =

    From Recall to Forgetting: Benchmarking Long-Term Memory for Personalized Agents , author =. 2026 , eprint =

    2026

  49. [57]

    2026 , url =

    Fang, Jizhan and Deng, Xinle and Xu, Haoming and Jiang, Ziyan and Tang, Yuqi and Xu, Ziwen and Deng, Shumin and Yao, Yunzhi and Wang, Mengru and Qiao, Shuofei and Chen, Huajun and Zhang, Ningyu , booktitle =. 2026 , url =

    2026