Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Federated Retrieval-Augmented Generation: A Systematic Mapping Study

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims to be the first systematic mapping study of Federated RAG, synthesizing 18 studies from 2020–2025 into a taxonomy of architectures, contributions, and application domains.

desk verdict A useful first sketch of a tiny field, but the missing search log and mislabeled corpus mean the 'first systematic mapping' claim doesn't hold up yet. read the letter →

arxiv 2505.18906 v2 pith:W3KNMKGY submitted 2025-05-24 cs.CL cs.IR

Abhijit Chakraborty , Chahana Dahal , Vivek Gupta This is my paper
classification cs.CLcs.IR
keywords federatedlearningretrieval-augmentedgenerationsystematicmappingstudyprivacy-preservingNLPretrievaltaxonomylargelanguagemodels
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims to be the first systematic mapping study of Federated Retrieval-Augmented Generation (RAG), a field that combines federated learning's privacy protections with retrieval-augmented generation's ability to ground language-model outputs in external knowledge. It assembles 18 primary studies published between 2020 and 2025 and classifies them by research focus, contribution type, and application domain, producing a taxonomy of federated RAG architectures. The mapping's value is that it turns a scattered set of system proposals into a comparable landscape: recurring design patterns, leading application domains, and a list of unsolved problems become visible at a glance. If the corpus and classification are accurate, the taxonomy is a reference tool for positioning future work and for deciding where the field needs benchmarks and evaluation protocols.

What carries the argument

The machinery that carries the argument is the classification scheme and architecture taxonomy (Figures 1 and 2), built on a manually coded corpus of 18 primary studies, and a unified objective function that expresses joint federated optimization of retrieval and generation. The taxonomy does the main analytical work: it reduces the field's design space to three axes—how retrieval is federated (naive, routed, or encrypted), how generation is placed (centralized, client-specific, or hybrid), and what privacy layer wraps the pipeline (enclaves, cryptography, or differential privacy)—which exposes both the dominant patterns and the empty cells. The objective function, $$L_{\mathrm{FedRAG}}(\$\theta$)=\sum_{i=1}^{M}\frac{n_i}{N}\left($L^{{(i)}}$_{\mathrm{retrieve}}(\theta_r)+$L^{{(i)}}$_{\mathrm{generate}}(\theta_r,\theta_g)\right),$$ formalizes the field's shared target: a single loss that balances local personalization against global consistency, making it clear what a complete Federated RAG system must optimize.

What would settle it

A search of the same venues and time window that produces an earlier systematic survey of Federated RAG, or an audit that finds one of the 18 included papers performs no retrieval augmentation, would overturn the 'first mapping study' and completeness claims.

Watch

Extended reading notes

Core claim

The paper's central claim is that it is the first structured synthesis of architectural, empirical, and deployment-oriented work in Federated RAG. Using a standard mapping-study methodology, the authors searched NLP, machine learning, and security venues from 2020 to 2025, identified 50 candidate papers, and screened these down to 18 primary studies. They coded each study by research focus (privacy and security, personalization and adaptation, retrieval efficiency, model integration), contribution type (new model or framework, dataset or benchmark, evaluation or analysis, survey or conceptual), and application domain (healthcare, finance and legal, enterprise and multilingual, recommendation and personal, general and multi-domain). The resulting taxonomy organizes the architecture space into three axes—federated retrieval (naive aggregation, selective query routing, secure index and query), federated generation (centralized, client-specific, hybrid), and privacy and security enhancements—and companion tables record the reported benefits of all 18 systems. The paper also introduces a unified objective function that frames retrieval and generation as jointly optimizable across clients.

Load-bearing premise

The load-bearing premise is that the 50 papers found and the 18 kept after screening are the complete, correctly labeled population of Federated RAG research published between 2020 and 2025.

Editorial extensions

If this is right

  • New systems can be positioned in the taxonomy immediately, letting authors state where their design sits on the retrieval, generation, and privacy axes.
  • The gap list gives concrete targets: scalable index synchronization, meta-learned retriever adaptation, privacy–utility Pareto benchmarks, and live benchmarks that track cost as documents drift.
  • The field's center of gravity—privacy and security in healthcare settings—is identifiable from the tables, so researchers can see where the crowded and the empty application areas are.
  • The conceptual objective function supplies a common optimization target for future work, even before any algorithm is shown to solve it.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 'first mapping study' claim is the most fragile part of the paper, because it rests on the assumption that no comparable survey appeared in the 2020–2025 window; a single missed earlier survey would reduce the claim to a contribution of synthesis rather than primacy.
  • Several of the 18 included works—PPDA, DP-FedKGE, DeTrustFL, FedX—are federated-learning or knowledge-graph systems that do not perform retrieval augmentation, so the true population of Federated RAG systems may be smaller than 18; the taxonomy's counts should be read with that inclusion bias in mind.
  • The taxonomy is likely reusable as a checklist for designing a new federated RAG system, but its axes may miss aspects the paper does not discuss, such as user-level personalization mechanisms or communication-constrained deployment; applying it to a new system would test its completeness.
  • A natural testable extension is to run the unified objective function on an existing benchmark like FeB4RAG to see whether joint optimization of retrieval and generation actually outperforms training them separately; the paper does not provide such evidence.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript reports a systematic mapping study of Federated RAG. It claims to be the first such study, covering 2020–2025, and bases its analysis on 18 primary studies selected from an initial pool of 50. The paper proposes a classification scheme, an architectural taxonomy, a set of application-domain categories, a unified objective function, and a gap analysis with future research directions. The Appendix provides a comparative synthesis and a list of benchmarks/toolkits.

Significance. Should the corpus be valid, the map would be a useful reference for an emerging field: it consolidates recent systems (C-FedRAG, FRAG, FedE4RAG, RAGRoute, MKP-QA, FeB4RAG), draws attention to evaluation gaps, and organizes design patterns. The choice of Kitchenham's methodology is appropriate in principle, and the authors are candid about limitations (possible missed studies, reproducibility concerns). However, the main contribution—a reliable, auditable map—is not achieved in the current version: the search protocol is absent, the inclusion decisions are inconsistent, and the numeric summaries do not reconcile with the claimed corpus. These problems affect the temporal trends, category counts, and gap analysis, so the positive elements do not yet amount to a defensible systematic mapping study.

major comments (4)
  1. [Section 2 and Limitations] No reproducible search or screening protocol is reported. The Methodology provides only "We searched the best NLP, ML and security venues (2020-2025) using terms such as ... including backward references," with no databases, query strings, inclusion/exclusion criteria, or screening log. A Kitchenham-style mapping study requires this protocol; without it the 18-paper corpus cannot be audited. The Limitations section itself concedes "limiting reproducibility," which is not a minor caveat here because every table and trend in the paper is descriptive of that corpus.
  2. [Section 3.1, Figure 1, Table 1] Several works counted as Federated RAG "Privacy & Security" studies do not appear to contain a retrieval-augmented generation component. PPDA (Jeon et al., 2021) is described as a decentralized aggregation protocol, DP-FedKGE (Peng et al., 2021) as a knowledge-graph embedding method, DeTrustFL (Xu et al., 2022) as a trust/consensus mechanism, and FedX (Han et al., 2022) as unsupervised representation learning. Including them inflates the Privacy & Security count (6) and the 2020–2022 foundational trend; excluding them would change the taxonomy and temporal narrative.
  3. [Sections 3.2–3.3, Tables 1–3, Appendix Table 5] The "18 primary studies" set is not well defined. Tables 1–3 sum to 13, 7, and 8, respectively; Figure 1 names 16 systems; Appendix Table 5 lists 18 entries but includes duplicates and conflations: "FRAG" and "FRAD (Zhao)" both cite Zhao 2024, "C-FedRAG" and "FedRAG" both cite Addison et al. 2024, and "RAG (Wang et al., 2024a)" appears alongside "UniMS-RAG (Wang et al., 2024a)". The distributions reported in the paper therefore cannot be verified from the supplied data.
  4. [Section 3.2 and Appendix A.1] The inclusion of RAGAS (Es et al., 2024) as a Federated RAG evaluation study is unsupported: it is cited as a general RAG evaluation toolkit, with no federated mechanism described. Similarly, MIRAGE (Xiong et al., 2024) is presented as a federated clinical benchmark, but the cited work is a medical RAG benchmark; no federated component is documented. Without an explicit inclusion criterion for what counts as Federated RAG, the evaluation-gap analysis and contribution-type counts are unreliable.
minor comments (5)
  1. [Section 3.1] "Personalisation" is used in Section 3.1 while "Personalization" is used elsewhere; please standardize the spelling.
  2. [Section 4] The mention of an "open-source FedRAG toolkit (Vector Institute)" is not accompanied by a citation, URL, or reference entry; please add the source or remove the claim.
  3. [Appendix A.2 and Section B] The "Full Comparative Synthesis of Surveyed Studies" appears twice in nearly identical form; the duplication should be removed.
  4. [Figure 4] The publication-trend figure reports counts but no table gives the per-year numeric values used; please provide the underlying data so the trend can be checked against the 18-study corpus.
  5. [Appendix A.1 and Reference List] The entry "FedRAG (Addison et al., 2024)" in Appendix A.1 has no matching reference entry; if it is intended to refer to Jiang (2024), the citation should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the mapping study's taxonomy and trends are descriptive classifications of an external literature corpus, not derivations from its own assumptions.

full rationale

The paper is a systematic mapping study, not a derivation. Its central claims are (a) that it is the first such mapping and (b) that the 18 surveyed studies can be organized by research focus, contribution type, and application domain. These claims are supported by external literature selection and classification, not by fitting parameters or by importing conclusions from the authors' own prior work. Section 3.4 introduces a 'conceptual objective' L_FedRAG(θ) = Σ (n_i/N)(L_retrieve + L_generate), but this is a standard weighted federated loss presented as a formalization, not as a prediction derived from the corpus; it is not used to infer any empirical claim. The Limitations section explicitly acknowledges that relevant studies may have been missed and that reproducibility is limited, but that is an external-validity concern, not circularity. There are no self-citations, no fitted inputs renamed as predictions, and no uniqueness theorem imported from the authors' own prior work. The internal inconsistencies in the reported counts (e.g., Tables 1–3 summing to 13, 7, and 8 rather than 18) are correctness and reporting risks, not circular reasoning. Therefore the derivation chain, such as it is, is self-contained and no circular step can be exhibited.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No new entities are postulated. The only equation (Section 3.4) is a standard weighted federated loss with ni/N weights and parameter sets theta_r, theta_g; these are definitions, not fitted values. The central contribution is a literature classification whose load-bearing assumptions are the completeness and correct labeling of the 18-study corpus, the applicability of Kitchenham's method, and the inclusion of federation-only papers as RAG-relevant.

assumptions (3)
  • domain assumption The 50 papers retrieved and 18 papers included constitute the relevant Federated RAG literature for 2020-2025.
    Section 2 describes searching 'the best NLP, ML and security venues' with backward references but provides no reproducible query, database list, or screening protocol. All RQ1-RQ4 findings depend on this corpus.
  • domain assumption Kitchenham's mapping-study guidelines were followed in a way that yields valid categories.
    Section 2 cites Kitchenham et al. (2011), but no quality assessment, data extraction form, or inter-rater reliability is reported; the resulting taxonomy is presented without validation.
  • ad hoc to paper Federated learning papers without retrieval-augmented generation (PPDA, DP-FedKGE, DeTrustFL, FedX) belong in the Federated RAG map.
    Section 3.1 classifies these as foundational privacy work, though they do not include retrieval or generation, broadening the field boundary to make the 18-study count possible.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Federated Retrieval-Augmented Generation: A Systematic Mapping Study." pith.science (2026). https://pith.science/paper/W3KNMKGY

@misc{pith2026250518906,
  author       = {Pith},
  title        = {Pith review of: Federated Retrieval-Augmented Generation: A Systematic Mapping Study},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W3KNMKGY}},
  note         = {Machine review of arXiv:2505.18906}
}
read the original abstract

Federated Retrieval-Augmented Generation (Federated RAG) combines Federated Learning (FL), which enables distributed model training without exposing raw data, with Retrieval-Augmented Generation (RAG), which improves the factual accuracy of language models by grounding outputs in external knowledge. As large language models are increasingly deployed in privacy-sensitive domains such as healthcare, finance, and personalized assistance, Federated RAG offers a promising framework for secure, knowledge-intensive natural language processing (NLP). To the best of our knowledge, this paper presents the first systematic mapping study of Federated RAG, covering literature published between 2020 and 2025. Following Kitchenham's guidelines for evidence-based software engineering, we develop a structured classification of research focuses, contribution types, and application domains. We analyze architectural patterns, temporal trends, and key challenges, including privacy-preserving retrieval, cross-client heterogeneity, and evaluation limitations. Our findings synthesize a rapidly evolving body of research, identify recurring design patterns, and surface open questions, providing a foundation for future work at the intersection of RAG and federated systems.

Figures

Figures reproduced from arXiv: 2505.18906 by the authors.

Figure 1
Figure 1. Classification Scheme: Summarizes the scheme with example categories and the distribution of studies (one study may fall into multiple categories). Federated RAG Architectures Federated Retrieval Naive Aggregation Selective Query Routing Secure Index & Query Federated Generation Centralized Generation Client-specific Generation Hybrid/Ensemble Generation Privacy/Security Enhancements Secure Execution Environments Cr… view at source ↗
Figure 2
Figure 2. Taxonomy of Federated RAG Architectures. A conceptual map to classify the diverse system designs in the literature. Research Focus # of Studies Privacy & Security 6 Personalization & Adaptation 2 Retrieval Efficiency 2 Model Integration 3 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Conceptual architecture of Federated RAG, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Publication trends on federated RAG from [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 25 canonical work pages

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Nguyen, Tomislav Medan, Jinali Shah, Mohammad T

    Parker Addison, Minh - Tuan H. Nguyen, Tomislav Medan, Jinali Shah, Mohammad T. Manzari, Brendan McElrone, Laksh Lalwani, Aboli More, Smita Sharma, Holger R. Roth, Isaac Yang, et al. 2024. C-FedRAG : A confidential federated retrieval-augmented generation system

    work page 2024

  4. [4]

    Brendan McMahan, Sarvar Patel, Daniel Ramage, Aaron Segal, and Karn Seth

    Keith Bonawitz, Vladimir Ivanov, Benjamin Kreuter, Alessandro Marcedone, H. Brendan McMahan, Sarvar Patel, Daniel Ramage, Aaron Segal, and Karn Seth. 2017. Practical secure aggregation for privacy-preserving machine learning. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security (CCS), pages 1175--1191

    work page 2017

  5. [5]

    Tom B Brown, Benjamin Mann, Nick Ryder, et al. 2020. Language models are few-shot learners. In NeurIPS, pages 1877--1901

    work page 2020

  6. [6]

    Tonmoy Debnath, Md Nurul Absar Siddiky, Muhammad Enayetur Rahman, Prosenjit Das, and Antu Kumar Guha. 2025. A comprehensive survey of prompt engineering techniques in large language models. TechRxiv

    work page 2025

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of NAACL-HLT

    work page 2019

  8. [8]

    Shahul Es, Jithin James, Luis Espinosa Anke, and Steven Schockaert. 2024. Ragas: Automated evaluation of retrieval augmented generation. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations, pages 150--158

    2024

Show all 34 references
  1. [9]

    Rachid Guerraoui, Anne-Marie Kermarrec, Diana Petrescu, Rafael Pires, Mathis Randl, and Martijn de Vos. 2025. Efficient federated search for retrieval-augmented generation. In Proceedings of the 5th Workshop on Machine Learning and Systems, pages 74--81

    2025

  2. [10]

    Sungwon Han, Sungwon Park, Fangzhao Wu, Sundong Kim, Chuhan Wu, Xing Xie, and Meeyoung Cha. 2022. Fedx: Unsupervised federated learning with cross knowledge distillation. In European Conference on Computer Vision, pages 691--707. Springer

    2022

  3. [11]

    Gautier Izacard and Edouard Grave. 2020. Leveraging passage retrieval with generative models for open domain question answering. arXiv preprint arXiv:2007.01282

    2020 arXiv

  4. [12]

    Beomyeol Jeon, S M Ferdous, Muntasir Raihan Rahman, and Anwar Walid. 2021. https://doi.org/10.1109/INFOCOMWKSHPS51825.2021.9484437 Privacy-preserving decentralized aggregation for federated learning . In IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops ...

    2021

  5. [13]

    Emily Jiang. 2024. Clinical Question-Answering over Distributed EHR Data. Ph.D. thesis, Massachusetts Institute of Technology

    2024

  6. [14]

    Jincheol Jung, Hongju Jeong, and Eui-Nam Huh. 2025. https://arxiv.org/abs/2412.13720 Federated learning and rag integration: A scalable approach for medical large language models . Preprint, arXiv:2412.13720

    2025 arXiv

  7. [15]

    Brendan McMahan, et al

    Peter Kairouz, H. Brendan McMahan, et al. 2021. Advances and open problems in federated learning. Foundations and Trends in Machine Learning, 14(1--2):1--210

    2021

  8. [16]

    Barbara Kitchenham, David Budgen, and Pearl Brereton. 2011. Mapping study methodology. Empirical Software Engineering, 16(6):791--810

    2011

  9. [17]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Yuxiang Ku, Wen-tau Chen, Guillaume Bouchard, Douwe Kiela, et al. 2020 a . Retrieval-augmented generation for knowledge-intensive nlp tasks. In Advances in Neural Information Process...

    2020

  10. [18]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Ilia Kulikov, Marjan Ghazvininejad, Wen - tau Yih, Tim Rockt \" a schel, et al. 2020 b . Retrieval-augmented generation for knowledge-intensive NLP tasks. In Advances in Neural Infor...

    2020

  11. [19]

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, et al. 2021. Pre-train prompt tune: Towards generalizing to unseen tasks. In Proceedings of ACL-IJCNLP

    2021

  12. [20]

    Qianren Mao, Qili Zhang, Hanwen Hao, Zhentao Han, Runhua Xu, Weifeng Jiang, Qi Hu, Zhijun Chen, Tyler Zhou, Bo Li, Yangqiu Song, Jin Dong, Jianxin Li, and Philip S. Yu. 2025. https://arxiv.org/abs/2504.19101 Privacy-preserving federated embedding learning for localized retriev...

    2025 arXiv

  13. [21]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas

    H. Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. In Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS)

    2017

  14. [22]

    Hao Peng, Haoran Li, Yangqiu Song, Vincent Zheng, and Jianxin Li. 2021. Differentially private federated knowledge graphs embedding. In Proceedings of the 30th ACM international conference on information & knowledge management, pages 1416--1425

    2021

  15. [23]

    Alec Radford, Jeffrey Wu, Rewon Child, et al. 2019. Language models are unsupervised multitask learners. OpenAI Blog

    2019

  16. [24]

    Timo Schick and Hinrich Schütze. 2021. It's not just size that matters: Small language models are also few-shot learners. In Proceedings of NAACL-HLT

    2021

  17. [25]

    Parshin Shojaee, Sai Sree Harsha, Dan Luo, Akash Maharaj, Tong Yu, and Yunyao Li. 2025. https://arxiv.org/abs/2501.14998 Federated retrieval augmented generation for multi-product question answering . Preprint, arXiv:2501.14998

    2025 arXiv

  18. [26]

    Milad Shokouhi and Luo Si. 2011. Federated search. Foundations and Trends in Information Retrieval, 5(1):1--102

    2011

  19. [27]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in neural information processing systems, pages 5998--6008

    2017

  20. [28]

    Pan, and Kam-Fai Wong

    Hongru Wang, Wenyu Huang, Yang Deng, Rui Wang, Zezhong Wang, Yufei Wang, Fei Mi, Jeff Z. Pan, and Kam-Fai Wong. 2024 a . https://arxiv.org/abs/2401.13256 Unims-rag: A unified multi-source retrieval-augmented generation for personalized dialogue systems . Preprint, arXiv:2401.13256

    2024 arXiv

  21. [29]

    Shuai Wang, Ekaterina Khramtsova, Shengyao Zhuang, and Guido Zuccon. 2024 b . Feb4rag: Evaluating federated search in the context of retrieval augmented generation. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retriev...

    2024

  22. [30]

    Yebo Wu, Chunlin Tian, Jingguang Li, He Sun, Kahou Tam, Li Li, and Chengzhong Xu. 2025. A survey on federated fine-tuning of large language models. arXiv preprint arXiv:2503.12016

    2025

  23. [31]

    Guangzhi Xiong, Qiao Jin, Zhiyong Lu, and Aidong Zhang. 2024. https://doi.org/10.18653/v1/2024.findings-acl.372 Benchmarking retrieval-augmented generation for medicine . In Findings of the Association for Computational Linguistics: ACL 2024, pages 6233--6251, Bangkok, Thailan...

    2024 doi

  24. [32]

    Runhua Xu, Nathalie Baracaldo, Yi Zhou, Ali Anwar, Swanand Kadhe, and Heiko Ludwig. 2022. Detrust-fl: Privacy-preserving federated learning in decentralized trust setting. In 2022 IEEE 15th International Conference on Cloud Computing (CLOUD), pages 417--426. IEEE

    2022

  25. [33]

    Huimin Zeng, Zhenrui Yue, Qian Jiang, and Dong Wang. 2024. Federated recommendation via hybrid retrieval augmented generation. In 2024 IEEE International Conference on Big Data (BigData), pages 8078--8087. IEEE

    2024

  26. [34]

    Dongfang Zhao. 2024. FRAG : Toward federated vector database management for collaborative and secure retrieval-augmented generation. arXiv preprint arXiv:2410.13272

    2024 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.