pith. machine review for the scientific record. sign in

arxiv: 2605.09794 · v1 · submitted 2026-05-10 · 💻 cs.IR

Recognition: no theorem link

LLM Agents Enable User-Governed Personalization Beyond Platform Boundaries

Jiacheng Lin , Kun Qian , Arvind Srinivasan , Tian Wang , Fang Han , Changran Hu , Junze Liu , Ziyi Wang
show 10 more authors
Hanwen Xu Mengmeng Xue Shuo Yang Hansi Zeng Simon Sinong Zhan Kai Zhong Weiqi Zhang Dakuo Wang Tianhao Wang Zhiyuan Li
Authors on Pith no claims yet

Pith reviewed 2026-05-12 02:34 UTC · model grok-4.3

classification 💻 cs.IR
keywords LLM agentspersonalizationcross-platform datauser data exportsrecommendation systemsdata integrationuser control
0
0 comments X

The pith

Users can outperform single-platform personalization by using off-the-shelf LLM agents to integrate their own cross-platform and offline data exports.

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

Personalization today stays fragmented because platforms cannot access complete user contexts due to competition, laws, and privacy limits. The paper shows that only users can collect their full data across services and real-world activities. LLM agents make this integration workable by reasoning over the mixed data formats to produce recommendations. Early tests demonstrate that this user-controlled method beats standard single-platform approaches. The work sets out a path for building systems that keep personalization under user direction.

Core claim

Large language model agents enable user-governed personalization by allowing users to aggregate and reason over their own heterogeneous data exports from multiple platforms plus offline sources, producing actionable outputs that exceed the capabilities of any individual platform's limited view.

What carries the argument

Off-the-shelf LLM agents that ingest and synthesize users' cross-platform data exports to generate integrated personalization without platform mediation.

If this is right

  • Users gain the ability to use their complete personal context for recommendations without sharing raw data with any service.
  • Personalization decisions move from platform algorithms to user-directed agent processes.
  • Data barriers between competing services become less restrictive for individual users.
  • New system designs can focus on agent reliability and safe data handling rather than platform data collection.

Where Pith is reading between the lines

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

  • Improved data export formats and APIs would make agent-based integration more reliable and widespread.
  • The model could incorporate offline personal records under strict user control, extending beyond digital traces.
  • Platform incentives might shift away from exclusive data ownership toward supporting user agents.

Load-bearing premise

Off-the-shelf LLM agents can reliably integrate and reason over heterogeneous personal data exports without significant errors or privacy leaks.

What would settle it

A user study in which recommendations generated by the LLM agent on real exported data score lower in relevance and accuracy than the outputs from the original single-platform models on matched tasks.

Figures

Figures reproduced from arXiv: 2605.09794 by Arvind Srinivasan, Changran Hu, Dakuo Wang, Fang Han, Hansi Zeng, Hanwen Xu, Jiacheng Lin, Junze Liu, Kai Zhong, Kun Qian, Mengmeng Xue, Shuo Yang, Simon Sinong Zhan, Tianhao Wang, Tian Wang, Weiqi Zhang, Zhiyuan Li, Ziyi Wang.

Figure 1
Figure 1. Figure 1: Platform-centric vs. user-governed personalization. Left: each platform observes only the behavioral data generated within its own service and personalizes in isolation. Right: the user aggregates data from multiple platforms and delegates personalization to an LLM agent, which reasons over the combined cross-platform context to produce recommendations. Although platforms are pursuing increasingly comprehe… view at source ↗
read the original abstract

Personalization today is fundamentally platform-centric: services build user representations from the behavioral fragments they observe. Yet no platform can construct a complete picture of the user, as competitive incentives, legal constraints, user privacy concerns, and epistemic limits create persistent data barriers. This paper argues for a shift from platform-centric personalization to user-governed personalization, where only the user can integrate fragmented contexts across platforms and the offline world. The key asymmetry lies in data access: only users can aggregate their own cross-platform and offline information. Large language model (LLM) agents make such integration practically feasible for the first time by enabling reasoning over heterogeneous personal data and transforming users' cross-context information into actionable personalization capabilities. We provide proof-of-concept evidence that users equipped with cross-platform data exports and an off-the-shelf LLM agent can outperform single-platform personalization baselines. We conclude by outlining a research agenda for building scalable user-governed personalization systems.

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 / 2 minor

Summary. The paper argues that personalization is inherently limited by platform-centric data collection due to competitive, legal, and privacy barriers, preventing any single service from building a complete user model. It proposes a shift to user-governed personalization, where individuals use their own cross-platform data exports (e.g., activity logs, notes, location histories) aggregated via off-the-shelf LLM agents to enable superior, context-aware personalization beyond any platform's reach. The central claim is supported by a proof-of-concept demonstration that such user-controlled LLM agents can outperform single-platform baselines, followed by an outline of a research agenda for scalable implementations.

Significance. If the proof-of-concept claim holds under rigorous testing, the work could meaningfully advance information retrieval and personalization research by reframing the problem around user-side data integration and LLM reasoning capabilities. It highlights an asymmetry in data access that platforms cannot replicate and identifies practical pathways for privacy-preserving cross-context systems. The explicit research agenda is a positive element, as it surfaces open questions in scalability, error handling, and system design that could guide follow-on empirical studies.

major comments (2)
  1. [Proof-of-Concept section] The proof-of-concept evidence for the headline claim (users with cross-platform exports and LLM agents outperforming single-platform baselines) is presented only at a high level without any description of the experimental protocol. No details are given on the downstream tasks evaluated, the specific data exports used, the prompting or agent architecture, the single-platform baselines, the quantitative metrics, or any error analysis for integration accuracy or hallucination rates. This absence directly undermines evaluation of the central performance claim.
  2. [Introduction and Proof-of-Concept] The manuscript does not provide any analysis or safeguards addressing the reliability of off-the-shelf LLM agents when fusing heterogeneous personal data (JSON logs, text, location histories). Without controlled measurements of factual errors, mis-integration, or data retention/leakage risks, the feasibility argument for user-governed personalization rests on an untested assumption that is load-bearing for the proposed asymmetry.
minor comments (2)
  1. [Introduction] The abstract and introduction would benefit from explicit citations to prior work on cross-platform data portability (e.g., GDPR data export studies) and LLM agent frameworks for personal data reasoning to better situate the contribution.
  2. [Introduction] Notation for 'user-governed personalization' is introduced without a formal definition or comparison table against related concepts such as federated personalization or personal data stores; a brief clarifying paragraph would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive comments. We agree that the proof-of-concept requires substantially more detail to support the central claims and that reliability considerations for LLM agents must be addressed explicitly. We will make major revisions to expand these sections accordingly while preserving the paper's focus on the conceptual shift to user-governed personalization.

read point-by-point responses

  1. Referee: [Proof-of-Concept section] The proof-of-concept evidence for the headline claim (users with cross-platform exports and LLM agents outperforming single-platform baselines) is presented only at a high level without any description of the experimental protocol. No details are given on the downstream tasks evaluated, the specific data exports used, the prompting or agent architecture, the single-platform baselines, the quantitative metrics, or any error analysis for integration accuracy or hallucination rates. This absence directly undermines evaluation of the central performance claim.

    Authors: We agree that the proof-of-concept is currently described at a high level without sufficient experimental details. This was a deliberate choice to keep the manuscript focused on the broader vision rather than a full empirical evaluation, but we recognize that it weakens the support for the performance claim. In the revised manuscript we will expand the Proof-of-Concept section with a complete description of the experimental protocol, including the specific downstream tasks (e.g., personalized search and recommendation scenarios), the concrete data exports used (e.g., JSON activity logs, text notes, and location histories), the off-the-shelf LLM agent architecture and prompting approach, the single-platform baselines, the quantitative metrics, and an error analysis covering integration accuracy and hallucination handling. revision: yes

  2. Referee: [Introduction and Proof-of-Concept] The manuscript does not provide any analysis or safeguards addressing the reliability of off-the-shelf LLM agents when fusing heterogeneous personal data (JSON logs, text, location histories). Without controlled measurements of factual errors, mis-integration, or data retention/leakage risks, the feasibility argument for user-governed personalization rests on an untested assumption that is load-bearing for the proposed asymmetry.

    Authors: We acknowledge that the manuscript does not currently include analysis or safeguards for LLM reliability when fusing heterogeneous personal data. The feasibility argument therefore relies on an assumption that requires explicit qualification. In the revision we will add a new subsection (either within Proof-of-Concept or as a dedicated Limitations discussion) that qualitatively analyzes known risks of factual errors, mis-integration, and data leakage, proposes practical user-side safeguards such as verification prompts and local processing, and notes that rigorous controlled measurements of these issues remain open questions to be addressed in the research agenda. revision: yes

Circularity Check

0 steps flagged

No circularity: conceptual argument with illustrative POC, no derivations or self-referential reductions

full rationale

The paper advances a conceptual position that only users can aggregate cross-platform data and that off-the-shelf LLM agents make such integration feasible, supported by a proof-of-concept. The provided text contains no equations, fitted parameters, derivations, or mathematical claims. No self-citations are used to justify uniqueness theorems or to smuggle in ansatzes. The central asymmetry argument and POC are presented as external evidence rather than results that reduce to the paper's own inputs by construction. This is a standard non-finding for a position paper without quantitative modeling.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the untested assumption that current LLMs can perform reliable integration of heterogeneous personal data; no free parameters or new physical entities are introduced.

axioms (1)
  • domain assumption LLM agents can effectively integrate and reason over heterogeneous personal data from multiple platforms and offline sources.
    Invoked to support the feasibility claim in the abstract.
invented entities (1)
  • user-governed personalization no independent evidence
    purpose: Framework that lets users aggregate and act on their own cross-platform data via LLM agents.
    New conceptual entity introduced to contrast with platform-centric approaches.

pith-pipeline@v0.9.0 · 5515 in / 1186 out tokens · 42288 ms · 2026-05-12T02:34:32.065522+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

50 extracted references · 50 canonical work pages · 2 internal anchors

  1. [1]

    Personalization technologies: a process- oriented perspective.Communications of the ACM, 48(10):83–90, 2005

    Gediminas Adomavicius and Alexander Tuzhilin. Personalization technologies: a process- oriented perspective.Communications of the ACM, 48(10):83–90, 2005

    work page 2005

  2. [2]

    Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions.IEEE transactions on knowledge and data engineering, 17(6):734–749, 2005

    Gediminas Adomavicius and Alexander Tuzhilin. Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions.IEEE transactions on knowledge and data engineering, 17(6):734–749, 2005

    work page 2005

  3. [3]

    Context-aware recommender systems

    Gediminas Adomavicius and Alexander Tuzhilin. Context-aware recommender systems. In Recommender systems handbook, pages 217–253. Springer, 2010

    work page 2010

  4. [4]

    Claude code overview

    Anthropic. Claude code overview. https://code.claude.com/docs/en/overview, n.d. Accessed: 2026-05-03

    work page 2026

  5. [5]

    Introducing apple intelligence for iphone, ipad, and mac

    Apple Inc. Introducing apple intelligence for iphone, ipad, and mac. https://www.apple.com/newsroom/2024/06/ introducing-apple-intelligence-for-iphone-ipad-and-mac/ , June 2024. Ac- cessed: 2026-05-03

    work page 2024

  6. [6]

    Americans and privacy: Concerned, confused and feeling lack of control over their personal information

    Brooke Auxier, Lee Rainie, Monica Anderson, Andrew Perrin, Madhu Kumar, and Erica Turner. Americans and privacy: Concerned, confused and feeling lack of control over their personal information. pew research center.Retrieved August, 22:2023, 2019

    work page 2023

  7. [7]

    Mediation of user models for en- hanced personalization in recommender systems.User Modeling and User-Adapted Interaction, 18(3):245–286, 2008

    Shlomo Berkovsky, Tsvi Kuflik, and Francesco Ricci. Mediation of user models for en- hanced personalization in recommender systems.User Modeling and User-Adapted Interaction, 18(3):245–286, 2008

    work page 2008

  8. [8]

    California consumer privacy act of 2018, cal

    California State Legislature. California consumer privacy act of 2018, cal. civ. code § 1798.100 et seq. https://leginfo.legislature.ca.gov/faces/codes_displayText.xhtml? division=3.&part=4.&lawCode=CIV&title=1.81.5, 2018. Accessed: 2026-05-03

    work page 2018

  9. [9]

    Cross- domain recommender systems

    Iván Cantador, Ignacio Fernández-Tobías, Shlomo Berkovsky, and Paolo Cremonesi. Cross- domain recommender systems. InRecommender systems handbook, pages 919–959. Springer, 2015

    work page 2015

  10. [10]

    Wide & deep learning for recommender systems

    Heng-Tze Cheng, Levent Koc, Jeremiah Harmsen, Tal Shaked, Tushar Chandra, Hrishi Aradhye, Glen Anderson, Greg Corrado, Wei Chai, Mustafa Ispir, et al. Wide & deep learning for recommender systems. InProceedings of the 1st workshop on deep learning for recommender systems, pages 7–10, 2016

    work page 2016

  11. [11]

    Deep neural networks for youtube recommenda- tions

    Paul Covington, Jay Adams, and Emre Sargin. Deep neural networks for youtube recommenda- tions. InProceedings of the 10th ACM conference on recommender systems, pages 191–198, 2016

    work page 2016

  12. [12]

    Commission finds apple and meta in breach of the digital mar- kets act

    European Commission. Commission finds apple and meta in breach of the digital mar- kets act. https://ec.europa.eu/commission/presscorner/detail/en/ip_25_1085, April 2025. Press release. Accessed: 2026-05-03

    work page 2025

  13. [13]

    European Parliament and Council of the European Union. Regulation (eu) 2016/679 of the european parliament and of the council of 27 april 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing directive 95/46/ec (general data protection regulation). https://eur-lex.euro...

    work page 2016

  14. [14]

    Regulation (eu) 2022/1925 of the european parliament and of the council of 14 september 2022 on contestable and fair markets in the digital sector (digital markets act)

    European Parliament and Council of the European Union. Regulation (eu) 2022/1925 of the european parliament and of the council of 14 september 2022 on contestable and fair markets in the digital sector (digital markets act). https://eur-lex.europa.eu/eli/reg/2022/ 1925/oj, October 2022. Official Journal of the European Union, L265, pp. 1–66. Accessed: 202...

    work page 2022

  15. [15]

    European Parliament and Council of the European Union. Regulation (eu) 2023/2854 of the european parliament and of the council of 13 december 2023 on harmonised rules on fair access to and use of data and amending regulation (eu) 2017/2394 and directive (eu) 2020/1828 (data act). https://eur-lex.europa.eu/eli/reg/2023/2854/oj, December 2023. Official Jour...

    work page 2023

  16. [16]

    Haiyan Fan and Marshall Scott Poole. What is personalization? perspectives on the design and implementation of personalization in information systems.Journal of Organizational Computing and Electronic Commerce, 16(3-4):179–202, 2006

    work page 2006

  17. [17]

    The interplay between the digital markets act and the general data protection regulation.Available at SSRN 4203907, 2022

    Damien Geradin, Konstantina Bania, and Theano Karanikioti. The interplay between the digital markets act and the general data protection regulation.Available at SSRN 4203907, 2022

    work page 2022

  18. [18]

    Using collaborative filtering to weave an information tapestry.Communications of the ACM, 35(12):61–70, 1992

    David Goldberg, David Nichols, Brian M Oki, and Douglas Terry. Using collaborative filtering to weave an information tapestry.Communications of the ACM, 35(12):61–70, 1992

    work page 1992

  19. [19]

    The role of artificial intelligence and data network effects for creating user value.Academy of management review, 46(3):534–551, 2021

    Robert Wayne Gregory, Ola Henfridsson, Evgeny Kaganer, and Harris Kyriakou. The role of artificial intelligence and data network effects for creating user value.Academy of management review, 46(3):534–551, 2021

    work page 2021

  20. [20]

    Bridging Language and Items for Retrieval and Recommendation: Benchmarking LLMs as Semantic Encoders

    Yupeng Hou, Jiacheng Li, Zhankui He, An Yan, Xiusi Chen, and Julian McAuley. Bridging language and items for retrieval and recommendation.arXiv preprint arXiv:2403.03952, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  21. [21]

    Large language models are zero-shot rankers for recommender systems

    Yupeng Hou, Junjie Zhang, Zihan Lin, Hongyu Lu, Ruobing Xie, Julian McAuley, and Wayne Xin Zhao. Large language models are zero-shot rankers for recommender systems. In European conference on information retrieval, pages 364–381. Springer, 2024

    work page 2024

  22. [22]

    Collaborative filtering for implicit feedback datasets

    Yifan Hu, Yehuda Koren, and Chris V olinsky. Collaborative filtering for implicit feedback datasets. In2008 Eighth IEEE international conference on data mining, pages 263–272. Ieee, 2008

    work page 2008

  23. [23]

    The challenge of understanding what users want: Inconsistent preferences and engagement optimization.Management science, 70(9):6336–6355, 2024

    Jon Kleinberg, Sendhil Mullainathan, and Manish Raghavan. The challenge of understanding what users want: Inconsistent preferences and engagement optimization.Management science, 70(9):6336–6355, 2024

    work page 2024

  24. [24]

    User modeling and user-adapted interaction

    Alfred Kobsa. User modeling and user-adapted interaction. InConference companion on Human factors in computing systems, pages 415–416, 1994

    work page 1994

  25. [25]

    Matrix factorization techniques for recom- mender systems.Computer, 42(8):30–37, 2009

    Yehuda Koren, Robert Bell, and Chris V olinsky. Matrix factorization techniques for recom- mender systems.Computer, 42(8):30–37, 2009

    work page 2009

  26. [26]

    Research in collaborative learning does not serve cross-silo federated learning in practice.arXiv preprint arXiv:2510.12595, 2025

    Kevin Kuo, Chhavi Yadav, and Virginia Smith. Research in collaborative learning does not serve cross-silo federated learning in practice.arXiv preprint arXiv:2510.12595, 2025

    work page arXiv 2025

  27. [27]

    Rec-r1: Bridging generative large language models and user-centric recommendation systems via reinforcement learning.Transactions on Machine Learning Research, 2025

    Jiacheng Lin, Tian Wang, and Kun Qian. Rec-r1: Bridging generative large language models and user-centric recommendation systems via reinforcement learning.Transactions on Machine Learning Research, 2025

    work page 2025

  28. [28]

    Greg Linden, Brent Smith, and Jeremy York. Amazon. com recommendations: Item-to-item collaborative filtering.IEEE Internet computing, 7(1):76–80, 2003

    work page 2003

  29. [29]

    Federated learning and free-riding in a competitive market.arXiv preprint arXiv:2410.12723, 2024

    Jiajun Meng, Jing Chen, Dongfang Zhao, and Lin Liu. Federated learning and free-riding in a competitive market.arXiv preprint arXiv:2410.12723, 2024

    work page arXiv 2024

  30. [30]

    Proof-of-concept for private local-to- cloud llm chat via trusted execution environments

    Avanika Narayan, Dan Biderman, and Christopher Re. Proof-of-concept for private local-to- cloud llm chat via trusted execution environments. InES-FoMo III: 3rd Workshop on Efficient Systems for Foundation Models, 2025

    work page 2025

  31. [31]

    Deep Learning Recommendation Model for Personalization and Recommendation Systems

    Maxim Naumov, Dheevatsa Mudigere, Hao-Jun Michael Shi, Jianyu Huang, Narayanan Sun- daraman, Jongsoo Park, Xiaodong Wang, Udit Gupta, Carole-Jean Wu, Alisson G Azzolini, et al. Deep learning recommendation model for personalization and recommendation systems. arXiv preprint arXiv:1906.00091, 2019. 12

    work page Pith review arXiv 1906

  32. [32]

    Nvidia announces dgx spark and dgx sta- tion personal ai computers

    NVIDIA Corporation. Nvidia announces dgx spark and dgx sta- tion personal ai computers. https://nvidianews.nvidia.com/news/ nvidia-announces-dgx-spark-and-dgx-station-personal-ai-computers , March

    work page

  33. [34]

    Introducing codex

    OpenAI. Introducing codex. https://openai.com/index/introducing-codex/, May

    work page

  34. [35]

    Accessed: 2026-05-03

    work page 2026

  35. [36]

    WW Norton & Company, 2016

    Geoffrey G Parker, Marshall W Van Alstyne, and Sangeet Paul Choudary.Platform revolution: How networked markets are transforming the economy and how to make them work for you. WW Norton & Company, 2016

    work page 2016

  36. [37]

    Google i/o 2025: Sundar pichai’s opening keynote

    Sundar Pichai. Google i/o 2025: Sundar pichai’s opening keynote. https://blog.google/ innovation-and-ai/technology/ai/io-2025-keynote/ , May 2025. Accessed: 2026- 05-03

    work page 2025

  37. [38]

    General data protection regulation.Intersoft Consulting, Accessed in October, 24(1), 2018

    Data Protection. General data protection regulation.Intersoft Consulting, Accessed in October, 24(1), 2018

    work page 2018

  38. [39]

    Competing with big data.The Journal of Industrial Economics, 69(4):967–1008, 2021

    Jens Prüfer and Christoph Schottmüller. Competing with big data.The Journal of Industrial Economics, 69(4):967–1008, 2021

    work page 2021

  39. [40]

    User modeling and user profiling: A comprehensive survey.arXiv preprint arXiv:2402.09660, 2024

    Erasmo Purificato, Ludovico Boratto, and Ernesto William De Luca. User modeling and user profiling: A comprehensive survey.arXiv preprint arXiv:2402.09660, 2024

    work page arXiv 2024

  40. [41]

    Grouplens: An open architecture for collaborative filtering of netnews

    Paul Resnick, Neophytos Iacovou, Mitesh Suchak, Peter Bergstrom, and John Riedl. Grouplens: An open architecture for collaborative filtering of netnews. InProceedings of the 1994 ACM conference on Computer supported cooperative work, pages 175–186, 1994

    work page 1994

  41. [42]

    Language representations can be what recommenders need: Findings and potentials

    Leheng Sheng, An Zhang, Yi Zhang, Yuxin Chen, Xiang Wang, and Tat-Seng Chua. Language representations can be what recommenders need: Findings and potentials. InThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  42. [43]

    Understanding the impact of web personalization on user information processing and decision outcomes1.MIS quarterly, 30(4):865–890, 2006

    Kar Yan Tam and Shuk Ying Ho. Understanding the impact of web personalization on user information processing and decision outcomes1.MIS quarterly, 30(4):865–890, 2006

    work page 2006

  43. [44]

    Api gover- nance: The case of facebook’s evolution.Social Media+ Society, 8(2):20563051221086228, 2022

    Fernando N Van der Vlist, Anne Helmond, Marcus Burkhardt, and Tatjana Seitz. Api gover- nance: The case of facebook’s evolution.Social Media+ Society, 8(2):20563051221086228, 2022

    work page 2022

  44. [45]

    A survey on large language models for recommendation

    Likang Wu, Zhi Zheng, Zhaopeng Qiu, Hao Wang, Hongchao Gu, Tingjia Shen, Chuan Qin, Chen Zhu, Hengshu Zhu, Qi Liu, et al. A survey on large language models for recommendation. World Wide Web, 27(5):60, 2024

    work page 2024

  45. [46]

    E-commerce product recommendation agents: Use, characteristics, and impact1.MIS quarterly, 31(1):137–209, 2007

    Bo Xiao and Izak Benbasat. E-commerce product recommendation agents: Use, characteristics, and impact1.MIS quarterly, 31(1):137–209, 2007

    work page 2007

  46. [47]

    Federated learning of gboard language models with differential privacy

    Zheng Xu, Yanxiang Zhang, Galen Andrew, Christopher Choquette, Peter Kairouz, Brendan Mcmahan, Jesse Rosenstock, and Yuanbo Zhang. Federated learning of gboard language models with differential privacy. InProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track), pages 629–639, 2023

    work page 2023

  47. [48]

    Qwen3 Technical Report

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  48. [49]

    Collm: Integrating collaborative embeddings into large language models for recommendation.IEEE Transactions on Knowledge and Data Engineering, 2025

    Yang Zhang, Fuli Feng, Jizhi Zhang, Keqin Bao, Qifan Wang, and Xiangnan He. Collm: Integrating collaborative embeddings into large language models for recommendation.IEEE Transactions on Knowledge and Data Engineering, 2025

    work page 2025

  49. [50]

    Zhehao Zhang, Ryan A. Rossi, Branislav Kveton, Yijia Shao, Diyi Yang, Hamed Zamani, Franck Dernoncourt, Joe Barrow, Tong Yu, Sungchul Kim, Ruiyi Zhang, Jiuxiang Gu, Tyler Derr, Hongjie Chen, Junda Wu, Xiang Chen, Zichao Wang, Subrata Mitra, Nedim Lipka, Nesreen K. Ahmed, and Yu Wang. Personalization of large language models: A survey. Transactions on Mach...

    work page 2025

  50. [51]

    Cross- domain recommendation: Challenges, progress, and prospects

    Feng Zhu, Yan Wang, Chaochao Chen, Jun Zhou, Longfei Li, and Guanfeng Liu. Cross- domain recommendation: Challenges, progress, and prospects. InProceedings of the Thirtieth International Joint Conference on Artificial Intelligence, pages 4721–4728. International Joint Conferences on Artificial Intelligence Organization, 2021. 14 A Ethics Statement This st...

    work page 2021