pith. sign in

arxiv: 2605.29640 · v2 · pith:HLKL4322new · submitted 2026-05-28 · 💻 cs.AI

VikingMem: A Memory Base Management System for Stateful LLM-based Applications

Jiajie Fu , Junwen Chen , Mengzhao Wang , Aoxiang He , Maojia Sheng , Xiangyu Ke , Yifan Zhu , Yunjun Gao This is my paper

Pith reviewed 2026-06-29 07:28 UTC · model grok-4.3

classification 💻 cs.AI
keywords Memory Basestateful LLM applicationsselective extractionstateful evolutiongeneralizable abstractiontemporal compressionmemory retrievalevent entity abstraction
0
0 comments X

The pith

VikingMem introduces a Memory Base paradigm that selectively extracts, evolves, and abstracts memories to manage persistent state in LLM interactions, outperforming baselines by up to 30 percent in retrieval effectiveness.

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

The paper proposes the Memory Base as a data management approach to address finite context windows that limit long-term state in LLM applications. It centers on three principles that extract high-value items from streams, allow memories to evolve through summarization and time-based adjustments, and support use across varied tasks without custom redesigns. VikingMem realizes the approach with events that drive entity updates, topic timelines for compression, and time-weighted recall to favor recent items. This setup is said to deliver stronger retrieval on benchmarks while preserving the speed needed for live user sessions. The evaluations focus on effectiveness gains relative to prior methods in memory handling.

Core claim

The paper claims that defining a Memory Base around selective extraction of high-value memories from raw streams, inherent statefulness and evolution through progressive summarization correction and temporal weighting, and generalizable abstraction for transfer across applications, then implementing it via event-centric extraction and dynamic entity updates with temporal compression on a topic-wise timeline and time-weighted recall, enables effective persistent state management for stateful LLM applications.

What carries the argument

The Memory Base paradigm defined by selective extraction, inherent statefulness and evolution, and generalizable abstraction, realized through interconnected event and entity abstractions together with topic-wise timeline compression and time-weighted recall.

If this is right

  • Memories are progressively summarized, corrected, and temporally weighted to prioritize recent interactions over older ones.
  • High-level summary memories are produced while older content is compressed and faded using time-weighted recall.
  • The approach transfers to downstream tasks including education, recommendation, and agent memory through its generalizable abstraction.
  • Low latency is preserved for interactive applications alongside the reported gains in retrieval effectiveness.

Where Pith is reading between the lines

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

  • The event-entity structure could support memory designs in multi-turn agent systems where state must persist across many turns.
  • Similar temporal weighting might be applied to other context-management techniques to reduce token usage in long sessions.
  • The abstraction layer suggests a path toward memory systems that avoid per-application prompt tuning.

Load-bearing premise

The three core principles of selective extraction, stateful evolution, and generalizable abstraction are sufficient to produce robust transferability across diverse downstream tasks without task-specific prompt engineering.

What would settle it

A controlled test on a new task such as medical consultation memory showing retrieval effectiveness no higher than baselines or a measurable rise in response latency would indicate the principles do not deliver the claimed general benefits.

Figures

Figures reproduced from arXiv: 2605.29640 by Aoxiang He, Jiajie Fu, Junwen Chen, Maojia Sheng, Mengzhao Wang, Xiangyu Ke, Yifan Zhu, Yunjun Gao.

Figure 1
Figure 1. Figure 1: The definition schema of event, entity for memory [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The pipeline of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The extract paradigm in VikingMem [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Use case for Agent Memory 4.2 Example Event-Entity Instance To provide a concrete understanding of how our abstract event￾entity model is applied in practice, we focus on one representative application scenario in greater detail. Specifically, we use Agent Memory to illustrate how raw interaction streams are transformed into structured event instances and further consolidated into persis￾tent entities. Mor… view at source ↗
read the original abstract

Large Language Models have revolutionized interactive applications; however, their finite context windows pose a critical data management challenge for maintaining stateful, long-term interactions. Existing memory approaches often rely on simplistic extraction methods that lead to incomplete memories or use rigid, single-purpose memory extraction prompts tailored to a single use case, such as chatbots. Consequently, they lack generalizability and perform poorly across diverse downstream tasks. To bridge this gap, we introduce the Memory Base, a novel data management paradigm for managing the persistent state of long-term interactions. It is characterized by three core principles: selective extraction of high-value memories from raw information streams; inherent statefulness and evolution, where memory content is progressively summarized, corrected, and temporally weighted to prioritize recent interactions; and a generalizable abstraction paradigm designed for robust transferability across diverse applications, including education, recommendation, and agent memory. Building on this foundation, we present VikingMem, an end-to-end Memory Base Management System implemented on the VikingDB vector engine. VikingMem materializes this paradigm through interconnected event and entity abstractions. It features event-centric memory extraction to selectively handle complex information streams, while entities are dynamically updated by events to achieve stateful evolution. Using temporal compression via a topic-wise timeline and time-weighted recall, the system progressively produces high-level summary memories, prioritizes recent items, and compresses and fades older ones. Extensive evaluations on long-term memory benchmarks demonstrate that VikingMem outperformes baselines by up to 30% in memory retrieval effectiveness while maintaining the low latency essential for interactive applications.

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 paper introduces the Memory Base paradigm for managing persistent state in long-term LLM interactions, defined by three principles: selective extraction of high-value memories, inherent statefulness with progressive summarization/correction/temporal weighting, and generalizable abstraction for transfer across tasks. It presents VikingMem, an implementation on VikingDB using interconnected event-centric extraction and entity updates, with topic-wise timelines and time-weighted recall for compression. The central claim is that extensive evaluations on long-term memory benchmarks show up to 30% improvement in retrieval effectiveness over baselines while preserving low latency for interactive use.

Significance. If the 30% retrieval gain is reproducible with named benchmarks, baselines, and error bars, and if the three principles demonstrably transfer to the stated domains (education, recommendation, agent memory), the work could offer a practical data-management layer for stateful LLM applications. The implementation details on VikingDB and the event/entity abstraction are potentially useful engineering contributions, but the absence of cross-domain results limits the assessed impact to incremental improvements on existing long-term memory tasks.

major comments (2)
  1. [Abstract] Abstract: The claim that the three core principles 'enable robust transferability across diverse applications, including education, recommendation, and agent memory' is not supported by the reported evaluations, which are described only as occurring on 'long-term memory benchmarks.' No results, ablations, or transfer experiments are mentioned for the explicitly named target domains, so the 30% figure cannot substantiate the broader system claim.
  2. [Abstract] Abstract: The performance claim of 'up to 30% in memory retrieval effectiveness' provides no benchmark names, baseline descriptions, statistical details, or error bars. Without these, it is impossible to assess whether the data support the central effectiveness claim or to compare against prior work.
minor comments (1)
  1. [Abstract] Abstract: Typo 'outperformes' should be 'outperforms'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We agree that the claims regarding transferability and the performance metric require qualification to align precisely with the reported evaluations, which focus on long-term memory benchmarks. We will revise the abstract accordingly in the next version.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the three core principles 'enable robust transferability across diverse applications, including education, recommendation, and agent memory' is not supported by the reported evaluations, which are described only as occurring on 'long-term memory benchmarks.' No results, ablations, or transfer experiments are mentioned for the explicitly named target domains, so the 30% figure cannot substantiate the broader system claim.

    Authors: We acknowledge that the evaluations are limited to long-term memory benchmarks and do not include explicit transfer experiments or results for the domains of education, recommendation, and agent memory. The abstract statement reflects the design intent of the Memory Base principles for generalizability, but we agree this is not empirically substantiated in the reported results. We will revise the abstract to state that the principles are designed to support transferability across applications, with empirical validation provided on long-term memory benchmarks relevant to these use cases. This change will be incorporated in the revised manuscript. revision: yes

  2. Referee: [Abstract] Abstract: The performance claim of 'up to 30% in memory retrieval effectiveness' provides no benchmark names, baseline descriptions, statistical details, or error bars. Without these, it is impossible to assess whether the data support the central effectiveness claim or to compare against prior work.

    Authors: The 30% figure is derived from the detailed experimental results in the manuscript body, which specify the long-term memory benchmarks, baselines, and include statistical measures such as error bars. However, we agree the abstract lacks sufficient context for standalone assessment. We will revise the abstract to qualify the claim (e.g., 'up to 30% on long-term memory benchmarks, as detailed in Section 5') or reference the key evaluation setup where space permits, ensuring the central claim is traceable to the full results without altering the reported numbers. revision: yes

Circularity Check

0 steps flagged

No circularity; system design and empirical claims rest on external benchmarks without self-referential reduction

full rationale

The paper presents a new Memory Base paradigm defined by three explicit principles (selective extraction, statefulness/evolution, generalizable abstraction) and an implementation (VikingMem) whose performance is asserted via evaluations on long-term memory benchmarks. No equations, fitted parameters, predictions, or self-citations appear in the abstract or described content that would make any claim equivalent to its inputs by construction. The transferability assertion to education/recommendation/agent domains is an untested claim rather than a derived result, but this constitutes an evidence gap, not circularity. The derivation chain is self-contained as a descriptive systems contribution evaluated against external baselines.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no visible free parameters, axioms, or invented entities; the central claim rests on unstated assumptions about benchmark representativeness and generalizability that cannot be audited from the provided text.

pith-pipeline@v0.9.1-grok · 5829 in / 1082 out tokens · 23437 ms · 2026-06-29T07:28:19.911730+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

80 extracted references · 24 canonical work pages · 11 internal anchors

  1. [1]

    Arkadeep Acharya, Brijraj Singh, and Naoyuki Onoe. 2023. Llm based generation of item-description for recommendation system. InProceedings of the 17th ACM conference on recommender systems. 1204–1207

    2023

  2. [2]

    Federico Barbero, Andrea Banino, Steven Kapturowski, Dharshan Kumaran, João Madeira Araújo, Oleksandr Vitvitskyi, Razvan Pascanu, and Petar Veličković

  3. [3]

    Advances in Neural Information Processing Systems37 (2024), 98111–98142

    Transformers need glasses! information over-squashing in language tasks. Advances in Neural Information Processing Systems37 (2024), 98111–98142

    2024

  4. [4]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners.Advances in neural information processing systems33 (2020), 1877–1901

    2020

  5. [5]

    Davide Caffagni, Federico Cocchi, Nicholas Moratelli, Sara Sarto, Marcella Cornia, Lorenzo Baraldi, and Rita Cucchiara. 2024. Wiki-llava: Hierarchical retrieval- augmented generation for multimodal llms. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. 1818–1826

    2024

  6. [6]

    Ningyuan Chen. 2023. Enterprises Build Sustainable Innovation Loop: Take ByteDance Launching Lark as an Example. InDigitalization and Management Innovation. IOS Press, 126–132

    2023

  7. [7]

    Sibei Chen, Ju Fan, Bin Wu, Nan Tang, Chao Deng, Pengyi Wang, Ye Li, Jian Tan, Feifei Li, Jingren Zhou, et al . 2025. Automatic database configuration debugging using retrieval-augmented language models.Proceedings of the ACM on Management of Data3, 1 (2025), 1–27

    2025

  8. [8]

    Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Ya- dav. 2025. Mem0: Building production-ready ai agents with scalable long-term memory.arXiv preprint arXiv:2504.19413(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  9. [9]

    Zhendong Chu, Shen Wang, Jian Xie, Tinghui Zhu, Yibo Yan, Jinheng Ye, Aoxiao Zhong, Xuming Hu, Jing Liang, Philip S Yu, et al. 2025. Llm agents for education: Advances and applications.arXiv preprint arXiv:2503.11733(2025)

    work page arXiv 2025

  10. [10]

    Dong Deng, Guoliang Li, Jianhua Feng, and Wen-Syan Li. 2013. Top-k string similarity search with edit-distance constraints. In2013 IEEE 29th International Conference on Data Engineering (ICDE). IEEE, 925–936

    2013

  11. [11]

    Weizhi Fei, Xueyan Niu, Pingyi Zhou, Lu Hou, Bo Bai, Lei Deng, and Wei Han

  12. [12]

    InFindings of the Association for Computational Linguistics ACL

    Extending Context Window of Large Language Models via Semantic Compression. InFindings of the Association for Computational Linguistics ACL

  13. [13]

    Jiajie Fu, Junwen Chen, Mengzhao Wang, Aoxiang He, Maojia Sheng, Xiangyu Ke, Yifan Zhu, and Yunjun Gao. 2025. Evaluation Code for VikingMem. https: //github.com/BytedanceFu/VikingMem

    2025

  14. [14]

    Jiajie Fu, Junwen Chen, Mengzhao Wang, Aoxiang He, Maojia Sheng, Xiangyu Ke, Yifan Zhu, and Yunjun Gao. 2025. Use cases for VikingMem. https://github. com/FuJiaJie123/VikingMem

    2025

  15. [15]

    Jiajie Fu, Haitong Tang, Arijit Khan, Sharad Mehrotra, Xiangyu Ke, and Yunjun Gao. 2025. In-context Clustering-based Entity Resolution with Large Language Models: A Design Space Exploration.arXiv preprint arXiv:2506.02509(2025)

    work page arXiv 2025

  16. [16]

    Aoran Gan, Hao Yu, Kai Zhang, Qi Liu, Wenyu Yan, Zhenya Huang, Shiwei Tong, and Guoping Hu. 2025. Retrieval Augmented Generation Evaluation in the Era of Large Language Models: A Comprehensive Survey.arXiv preprint arXiv:2504.14891(2025)

    work page arXiv 2025

  17. [17]

    Jianyang Gao and Cheng Long. 2024. Rabitq: Quantizing high-dimensional vectors with a theoretical error bound for approximate nearest neighbor search. Proceedings of the ACM on Management of Data2, 3 (2024), 1–27

    2024

  18. [18]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. 2024. Retrieval-Augmented Generation for Large Language Models: A Survey. arXiv:2312.10997 [cs.CL] https://arxiv.org/abs/2312.10997

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  19. [19]

    Julie Gonneaud, Géraldine Rauchs, Mathilde Groussard, Brigitte Landeau, Flo- rence Mezenge, Vincent de La Sayette, Francis Eustache, and Béatrice Desgranges

  20. [20]

    How do we process event-based and time-based intentions in the brain? An fMRI study of prospective memory in healthy individuals.Human brain mapping35, 7 (2014), 3066–3082

    2014

  21. [21]

    Gaurav Goswami. 2025. Dissecting the metrics: How different evaluation ap- proaches yield diverse results for conversational ai.Authorea Preprints(2025)

    2025

  22. [22]

    Qinyu Han, Zhihao Yang, Hongfei Lin, and Tian Qin. 2024. Let topic flow: A uni- fied topic-guided segment-wise dialogue summarization framework.IEEE/ACM Transactions on Audio, Speech, and Language Processing32 (2024), 2021–2032

    2024

  23. [23]

    Xiaoxin He, Yijun Tian, Yifei Sun, Nitesh Chawla, Thomas Laurent, Yann LeCun, Xavier Bresson, and Bryan Hooi. 2024. G-retriever: Retrieval-augmented gen- eration for textual graph understanding and question answering.Advances in Neural Information Processing Systems37 (2024), 132876–132907

    2024

  24. [24]

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, et al. 2025. A survey on hallucination in large language models: Principles, taxonomy, chal- lenges, and open questions.ACM Transactions on Information Systems43, 2 (2025), 1–55

    2025

  25. [25]

    Yiqian Huang, Shiqi Zhang, and Xiaokui Xiao. 2025. KET-RAG: A Cost-Efficient Multi-Granular Indexing Framework for Graph-RAG. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining(Toronto ON, Canada)(KDD ’25). Association for Computing Machinery, 1003–1012

    2025

  26. [26]

    Brennan Jones, Yan Xu, Qisheng Li, and Stefan Scherer. 2024. Designing a Proactive Context-Aware AI Chatbot for People’s Long-Term Goals. InExtended Abstracts of the CHI Conference on Human Factors in Computing Systems. 1–7

    2024

  27. [27]

    Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick SH Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense Passage Retrieval for Open-Domain Question Answering.. InEMNLP (1). 6769–6781

    2020

  28. [28]

    Omar Khattab and Matei Zaharia. 2020. Colbert: Efficient and effective passage search via contextualized late interaction over bert. InProceedings of the 43rd International ACM SIGIR conference on research and development in Information Retrieval. 39–48

    2020

  29. [29]

    Andrey Kuzmin, Mart Van Baalen, Yuwei Ren, Markus Nagel, Jorn Peters, and Ti- jmen Blankevoort. 2022. Fp8 quantization: The power of the exponent.Advances in Neural Information Processing Systems35 (2022), 14651–14662

    2022

  30. [30]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rock- täschel, et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks.Advances in neural information processing systems33 (2020), 9459–9474

    2020

  31. [31]

    Hao Li, Chenghao Yang, An Zhang, Yang Deng, Xiang Wang, and Tat-Seng Chua

  32. [32]

    In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)

    Hello again! llm-powered personalized agent for long-term dialogue. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 5259–5276

    2025

  33. [33]

    Xinze Li, Yushi Bai, Bowen Jin, Fengbin Zhu, Liangming Pan, and Yixin Cao

  34. [34]

    RAG: Strategies for Processing Long Documents in LLMs

    Long Context vs. RAG: Strategies for Processing Long Documents in LLMs. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. 4110–4113

  35. [35]

    Yuhong Li, Yingbing Huang, Bowen Yang, Bharat Venkitesh, Acyr Locatelli, Hanchen Ye, Tianle Cai, Patrick Lewis, and Deming Chen. 2024. Snapkv: Llm knows what you are looking for before generation.Advances in Neural Informa- tion Processing Systems37 (2024), 22947–22970

    2024

  36. [36]

    Zhiyu Li, Shichao Song, Hanyu Wang, Simin Niu, Ding Chen, Jiawei Yang, Chenyang Xi, Huayi Lai, Jihao Zhao, Yezhaohui Wang, et al. 2025. MemOS: An Operating System for Memory-Augmented Generation (MAG) in Large Language Models.arXiv preprint arXiv:2505.22101(2025)

    work page arXiv 2025

  37. [37]

    Patricia W Linville. 1987. Self-complexity as a cognitive buffer against stress- related illness and depression.Journal of personality and social psychology52, 4 (1987), 663

    1987

  38. [38]

    Jiaheng Liu, Dawei Zhu, Zhiqi Bai, Yancheng He, Huanxuan Liao, Haoran Que, Zekun Wang, Chenchen Zhang, Ge Zhang, Jiebin Zhang, et al. 2025. A comprehen- sive survey on long context language modeling.arXiv preprint arXiv:2503.17407 (2025)

    work page arXiv 2025

  39. [39]

    Hanjia Lyu, Song Jiang, Hanqing Zeng, Yinglong Xia, Qifan Wang, Si Zhang, Ren Chen, Chris Leung, Jiajie Tang, and Jiebo Luo. 2024. Llm-rec: Personal- ized recommendation via prompting large language models. InFindings of the Association for Computational Linguistics: NAACL 2024. 583–612

    2024

  40. [40]

    Xinbei Ma, Yeyun Gong, Pengcheng He, Hai Zhao, and Nan Duan. 2023. Query rewriting in retrieval-augmented large language models. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 5303–5315

    2023

  41. [41]

    Yubo Ma, Jinsong Li, Yuhang Zang, Xiaobao Wu, Xiaoyi Dong, Pan Zhang, Yuhang Cao, Haodong Duan, Jiaqi Wang, Yixin Cao, and Aixin Sun. 2025. To- wards Storage-Efficient Visual Document Retrieval: An Empirical Study on Re- ducing Patch-Level Embeddings. InFindings of the Association for Computational Linguistics: ACL 2025. 19568–19580

    2025

  42. [42]

    Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. 2024. Evaluating Very Long-Term Conversational Memory of LLM Agents. InProceedings of the 62nd Annual Meeting of the Associ- ation for Computational Linguistics (Volume 1: Long Papers). 13851–13870

    2024

  43. [43]

    Hazel Markus. 1977. Self-schemata and processing information about the self. Journal of personality and social psychology35, 2 (1977), 63. 13

    1977

  44. [44]

    Memobase. 2025. Source code of Memobase. https://github.com/memodb- io/memobase

    2025

  45. [45]

    Mite Mijalkov, Ludvig Storm, Blanca Zufiria-Gerbolés, Dániel Veréb, Zhilei Xu, Anna Canal-Garcia, Jiawei Sun, Yu-Wei Chang, Hang Zhao, Emiliano Gómez- Ruiz, et al. 2025. Computational memory capacity predicts aging and cognitive decline.Nature communications16, 1 (2025), 2748

    2025

  46. [46]

    Openclaw. 2026. Openclaw AI. https://openclaw.ai/

    2026

  47. [47]

    Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G Patil, Ion Stoica, and Joseph E Gonzalez. 2023. MemGPT: Towards LLMs as Operating Systems.arXiv preprint arXiv:2310.08560(2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  48. [48]

    Zaifeng Pan, Ajjkumar Patel, Zhengding Hu, Yipeng Shen, Yue Guan, Wan-Lu Li, Lianhui Qin, Yida Wang, and Yufei Ding. 2025. KVFlow: Efficient Prefix Caching for Accelerating LLM-Based Multi-Agent Workflows.arXiv preprint arXiv:2507.07400(2025)

    work page arXiv 2025

  49. [49]

    Vicky Zhao, Lili Qiu, and Jianfeng Gao

    Zhuoshi Pan, Qianhui Wu, Huiqiang Jiang, Xufang Luo, Hao Cheng, Dongsheng Li, Yuqing Yang, Chin-Yew Lin, H. Vicky Zhao, Lili Qiu, and Jianfeng Gao. 2025. SeCom: On Memory Construction and Retrieval for Personalized Conversational Agents. InThe Thirteenth International Conference on Learning Representations. https://openreview.net/forum?id=xKDZAW0He3

    2025

  50. [50]

    Bowen Peng, Jeffrey Quesnelle, Honglu Fan, and Enrico Shippole. 2023. Yarn: Efficient context window extension of large language models.arXiv preprint arXiv:2309.00071(2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  51. [51]

    Preston Rasmussen, Pavlo Paliychuk, Travis Beauvais, Jack Ryan, and Daniel Chalef. 2025. Zep: a temporal knowledge graph architecture for agent memory. arXiv preprint arXiv:2501.13956(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  52. [52]

    Machel Reid et al. 2024. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context.arXiv preprint arXiv:2403.05530(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  53. [53]

    Henry L Roediger III and Jeffrey D Karpicke. 2006. Test-enhanced learning: Taking memory tests improves long-term retention.Psychological science17, 3 (2006), 249–255

    2006

  54. [54]

    RooCodeInc. 2026. Roocode Source Code. Retrieved January 11, 2026 from https://github.com/RooCodeInc/Roo-Code

    2026

  55. [55]

    Vitória S Santos and Carina F Dorneles. 2024. Unveiling the Segmentation Power of LLMs: Zero-Shot Invoice Item Description Analysis. InSimpósio Brasileiro de Banco de Dados (SBBD). SBC, 549–561

    2024

  56. [56]

    Chaitanya Sharma. 2025. Retrieval-Augmented Generation: A Comprehensive Survey of Architectures, Enhancements, and Robustness Frontiers.arXiv preprint arXiv:2506.00054(2025)

    work page arXiv 2025

  57. [57]

    Yunxiao Shi, Xing Zi, Zijing Shi, Haimin Zhang, Qiang Wu, and Min Xu. 2024. En- hancing Retrieval and Managing Retrieval: A Four-Module Synergy for Improved Quality and Efficiency in RAG Systems. InECAI 2024. IOS Press, 2258–2265

    2024

  58. [58]

    Aditi Singh, Abul Ehtesham, Saket Kumar, and Tala Talaei Khoei. 2025. Agen- tic retrieval-augmented generation: A survey on agentic rag.arXiv preprint arXiv:2501.09136(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  59. [59]

    Weihang Su, Yichen Tang, Qingyao Ai, Junxi Yan, Changyue Wang, Hongning Wang, Ziyi Ye, Yujia Zhou, and Yiqun Liu. 2025. Parametric retrieval augmented generation. InProceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. 1240–1250

    2025

  60. [60]

    Jakub Swacha and Michał Gracel. 2025. Retrieval-Augmented Generation (RAG) Chatbots for Education: A Survey of Applications.Applied Sciences15, 8 (2025), 4234

    2025

  61. [61]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971(2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  62. [62]

    VikingMem. 2025. VikingMem User Guide. https://docs.byteplus.com/en/docs/ VikingDB/Integrating_LLM_with_Memory

    2025

  63. [63]

    Xi Wang, Procheta Sen, Ruizhe Li, and Emine Yilmaz. 2024. Adaptive Retrieval-Augmented Generation for Conversational Systems.arXiv preprint arXiv:2407.21712(2024)

    work page arXiv 2024

  64. [64]

    Yu Wang and Xi Chen. 2025. MIRIX: Multi-Agent Memory System for LLM-Based Agents.arXiv preprint arXiv:2507.07957(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  65. [65]

    Zora Zhiruo Wang, Jiayuan Mao, Daniel Fried, and Graham Neubig. 2024. Agent workflow memory.arXiv preprint arXiv:2409.07429(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  66. [66]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reason- ing in large language models.Advances in neural information processing systems 35 (2022), 24824–24837

    2022

  67. [67]

    Chunlong Wu, Ye Luo, Zhibo Qu, and Min Wang. 2025. Meta-Policy Reflexion: Reusable Reflective Memory and Rule Admissibility for Resource-Efficient LLM Agent.arXiv preprint arXiv:2509.03990(2025)

    work page arXiv 2025

  68. [68]

    Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. 2024. Longmemeval: Benchmarking chat assistants on long-term interactive memory.arXiv preprint arXiv:2410.10813(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  69. [69]

    Weiqi Wu, Shen Huang, Yong Jiang, Pengjun Xie, Fei Huang, and Hai Zhao

  70. [70]

    InFindings of the Association for Computational Linguistics: NAACL 2025

    Unfolding the Headline: Iterative Self-Questioning for News Retrieval and Timeline Summarization. InFindings of the Association for Computational Linguistics: NAACL 2025. 4385–4398

    2025

  71. [71]

    Yunzhong Xiao, Yangmin Li, Hewei Wang, Yunlong Tang, and Zora Zhiruo Wang

  72. [72]

    ToolMem: Enhancing Multimodal Agents with Learnable Tool Capability Memory.arXiv preprint arXiv:2510.06664(2025)

    work page arXiv 2025

  73. [73]

    Linhao Ye, Zhikai Lei, Jianghao Yin, Qin Chen, Jie Zhou, and Liang He

  74. [74]

    InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval

    Boosting conversational question answering with fine-grained retrieval- augmentation and self-check. InProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. 2301–2305

  75. [75]

    Haoyu Zhang, Jun Liu, Zhenhua Zhu, Shulin Zeng, Maojia Sheng, Tao Yang, Guohao Dai, and Yu Wang. 2024. Efficient and effective retrieval of dense-sparse hybrid vectors using graph-based approximate nearest neighbor search.arXiv preprint arXiv:2410.20381(2024)

    work page arXiv 2024

  76. [76]

    Xiang Zhang, Juntai Cao, and Chenyu You. 2024. Counting ability of large language models and impact of tokenization.arXiv preprint arXiv:2410.19730 (2024)

    work page arXiv 2024

  77. [77]

    Xin Zhang, Yanzhao Zhang, Dingkun Long, Wen Xie, Ziqi Dai, Jialong Tang, Huan Lin, Baosong Yang, Pengjun Xie, Fei Huang, et al. 2024. mGTE: Generalized Long-Context Text Representation and Reranking Models for Multilingual Text Retrieval. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track. 1393–1412

    2024

  78. [78]

    Siyun Zhao, Yuqing Yang, Zilong Wang, Zhiyuan He, Luna K Qiu, and Lili Qiu

  79. [79]

    Retrieval augmented generation (rag) and beyond: A comprehensive survey on how to make your llms use external data more wisely.arXiv preprint arXiv:2409.14924(2024)

    work page arXiv 2024

  80. [80]

    Wei Zhou, Yuyang Gao, Xuanhe Zhou, and Guoliang Li. 2025. Cracking SQL Barriers: An LLM-based Dialect Translation System.Proceedings of the ACM on Management of Data3, 3 (2025), 1–26. 14

    2025