pith. machine review for the scientific record. sign in

arxiv: 2604.10678 · v1 · submitted 2026-04-12 · 💻 cs.AI · cs.LG

Recognition: unknown

FedRio: Personalized Federated Social Bot Detection via Cooperative Reinforced Contrastive Adversarial Distillation

Yingguang Yang , Hao Liu , Xin Zhang , Yunhui Liu , Yutong Xia , Qi Wu , Hao Peng , Taoran Liang
show 3 more authors
Bin Chong Tieke He Philip S. Yu
Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:11 UTC · model grok-4.3

classification 💻 cs.AI cs.LG
keywords federated learningsocial bot detectioncontrastive learningadversarial distillationreinforcement learninggraph neural networkspersonalized federated learningprivacy preservation
0
0 comments X

The pith

FedRio lets social platforms detect bots together without sharing user data by using GANs, contrastive learning, and reinforcement control to handle differences across clients.

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

The paper develops a federated framework called FedRio to let separate social platforms improve their bot detectors by sharing patterns while keeping all raw data local. It tackles the core problems of mismatched data distributions, differing model architectures, and privacy rules that block simple central training. The method adds an adaptive graph message-passing layer on each client, uses generative adversarial networks to extract shared knowledge, applies multi-stage adversarial contrastive training to align feature spaces, and employs reinforcement learning to tune client parameters during aggregation. Experiments on two real-world bot detection datasets show gains in accuracy and communication cost over other federated baselines, with results close to fully centralized models despite the added privacy constraints.

Core claim

FedRio shows that an adaptive message-passing graph neural network backbone combined with GAN-based federated knowledge extraction, multi-stage adversarial contrastive learning, and reinforcement learning-based client parameter control produces a personalized federated bot detector that outperforms prior federated methods in accuracy, communication efficiency, and feature-space alignment on two public benchmarks while staying competitive with published centralized detectors under stricter privacy rules.

What carries the argument

The cooperative reinforced contrastive adversarial distillation process, which uses GANs to transfer global distribution knowledge, adversarial contrastive objectives to reduce local-global divergence, and reinforcement learning to adaptively control client contributions during server aggregation.

Load-bearing premise

The modules for message passing, GAN knowledge extraction, adversarial contrastive alignment, and reinforcement learning parameter control will reliably manage client data and architecture differences without training instability or excessive communication overhead.

What would settle it

On the two real-world social bot detection benchmarks, if FedRio fails to exceed the accuracy or communication efficiency of the strongest federated baselines reported in the paper, or if feature-space consistency metrics do not improve, the central performance claims would be falsified.

Figures

Figures reproduced from arXiv: 2604.10678 by Bin Chong, Hao Liu, Hao Peng, Philip S. Yu, Qi Wu, Taoran Liang, Tieke He, Xin Zhang, Yingguang Yang, Yunhui Liu, Yutong Xia.

Figure 1
Figure 1. Figure 1: The proposed FEDRIO framework. of a GNN backbone and a fully connected layer. The GNN dynamically adjusts the message propagation for each node to enable personalized node representations. The overall objective is to minimize the total error across all clients. The server does not collect raw data from clients but aggregates model parameters to tackle challenges arising from non-identically distributed dat… view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of data heterogeneity. The darker color means more training samples with a label available to the client. [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Loss dynamics during training under different data consistency parameters and loss functions. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Hyperparameter Sensitivity (𝛾, 𝜏, 𝜇) of FedACK on Vendor-19 under diferent data heterogeneity settings (𝛼). FedAvg FedProx Ensemble FedDistill FedGen FedACK FedRio w/o amp FedRio 0 25 50 75 100 Communication Rounds 0.2 0.4 0.6 0.8 1.0 Accuracy (a) Vendor-19 (𝛼 = 1). 0 25 50 75 100 Communication Rounds 0.2 0.4 0.6 0.8 1.0 Accuracy (b) Vendor-19 (𝛼 = 0.5). 0 25 50 75 100 Communication Rounds 0.45 0.55 0.65 0… view at source ↗
Figure 5
Figure 5. Figure 5: Learning Curve of (a-b) Vendor-19 and (c-d) TwiBot-20 in 100 communication rounds in different [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Decision boundaries and feature space of two randomly selected [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
read the original abstract

Social bot detection is critical to the stability and security of online social platforms. However, current state-of-the-art bot detection models are largely developed in isolation, overlooking the benefits of leveraging shared detection patterns across platforms to improve performance and promptly identify emerging bot variants. The heterogeneity of data distributions and model architectures further complicates the design of an effective cross-platform and cross-model detection framework. To address these challenges, we propose FedRio (Personalized Federated Social Bot Detection with Cooperative Reinforced Contrastive Adversarial Distillation framework. We first introduce an adaptive message-passing module as the graph neural network backbone for each client. To facilitate efficient knowledge sharing of global data distributions, we design a federated knowledge extraction mechanism based on generative adversarial networks. Additionally, we employ a multi-stage adversarial contrastive learning strategy to enforce feature space consistency among clients and reduce divergence between local and global models. Finally, we adopt adaptive server-side parameter aggregation and reinforcement learning-based client-side parameter control to better accommodate data heterogeneity in heterogeneous federated settings. Extensive experiments on two real-world social bot detection benchmarks demonstrate that FedRio consistently outperforms state-of-the-art federated learning baselines in detection accuracy, communication efficiency, and feature space consistency, while remaining competitive with published centralized results under substantially stronger privacy constraints.

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

3 major / 2 minor

Summary. The paper proposes FedRio, a personalized federated learning framework for cross-platform social bot detection. It introduces an adaptive message-passing GNN backbone per client, a GAN-based federated knowledge extraction mechanism, a multi-stage adversarial contrastive learning strategy to enforce feature consistency, and adaptive server-side aggregation combined with RL-based client-side parameter control to handle data and architectural heterogeneity. Extensive experiments on two real-world benchmarks are claimed to show consistent outperformance over state-of-the-art federated baselines in detection accuracy, communication efficiency, and feature space consistency, while remaining competitive with centralized methods under stronger privacy constraints.

Significance. If the empirical claims hold under rigorous validation, the work would be significant for advancing privacy-preserving machine learning in social network security, where cross-platform collaboration is needed to detect emerging bot variants without sharing raw data. The combination of GAN distillation, adversarial contrastive alignment, and RL-driven adaptation for heterogeneous clients offers a concrete approach to personalized federated learning that could inform applications beyond bot detection. No machine-checked proofs or parameter-free derivations are present, but the reproducible experimental setup on public benchmarks would be a strength if ablations and stability analyses are added.

major comments (3)
  1. [§4] §4 (Experiments): The abstract and results claim consistent outperformance and robustness to heterogeneity, but no convergence curves, ablation tables removing the GAN, contrastive, or RL components, or breakdowns by heterogeneity level are referenced; without these, it is impossible to confirm that the multi-component design does not trade stability for the reported gains in accuracy and efficiency.
  2. [§3.2–3.4] §3.2–3.4 (Method): The GAN-based knowledge extraction and multi-stage adversarial contrastive learning are presented as solutions to distribution divergence, yet the manuscript provides no analysis of potential mode collapse in the GAN or oscillation in the RL policy under realistic client heterogeneity; this directly bears on whether the claimed feature space consistency and communication savings are achievable.
  3. [Table 2] Table 2 or main results table: Reported improvements over federated baselines lack statistical significance tests, standard deviations across multiple runs, or hyperparameter sensitivity analysis; this undermines the central claim that FedRio reliably outperforms under stronger privacy constraints.
minor comments (2)
  1. [§3.4] Notation for the RL reward function and the contrastive loss weighting schedule should be clarified with explicit equations to avoid ambiguity in reproduction.
  2. [Figure 3] Figure captions for the architecture diagram and feature space visualizations could include more detail on what the t-SNE plots specifically demonstrate regarding client alignment.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback, as well as the recommendation for major revision. We address each major comment point by point below and will incorporate the requested additions and analyses into the revised manuscript to strengthen the empirical validation and statistical rigor of our claims.

read point-by-point responses

  1. Referee: [§4] §4 (Experiments): The abstract and results claim consistent outperformance and robustness to heterogeneity, but no convergence curves, ablation tables removing the GAN, contrastive, or RL components, or breakdowns by heterogeneity level are referenced; without these, it is impossible to confirm that the multi-component design does not trade stability for the reported gains in accuracy and efficiency.

    Authors: We agree that these elements are essential for rigorously validating the contributions of each component and the overall stability. In the revised manuscript, we will add convergence curves for key metrics (accuracy, communication cost, and feature consistency) across training rounds for FedRio and the main baselines. We will also include ablation tables that systematically remove the GAN-based knowledge extraction, the multi-stage adversarial contrastive learning, and the RL-based client-side control, reporting the resulting performance drops. Finally, we will provide breakdowns of results stratified by increasing levels of data heterogeneity (measured via distribution divergence metrics) and architectural heterogeneity (e.g., varying GNN depths and message-passing variants per client). These additions will directly address concerns about potential stability-accuracy trade-offs. revision: yes

  2. Referee: [§3.2–3.4] §3.2–3.4 (Method): The GAN-based knowledge extraction and multi-stage adversarial contrastive learning are presented as solutions to distribution divergence, yet the manuscript provides no analysis of potential mode collapse in the GAN or oscillation in the RL policy under realistic client heterogeneity; this directly bears on whether the claimed feature space consistency and communication savings are achievable.

    Authors: We acknowledge that the current manuscript lacks an explicit stability analysis for these components. In the revision, we will add a new subsection in the experiments (with supporting discussion in the method) that empirically examines GAN training dynamics, including metrics for sample diversity to detect mode collapse, and RL policy stability, including variance in policy gradients and reward convergence under varying client heterogeneity levels. We will report that our runs exhibited stable behavior without collapse or excessive oscillation, supported by plots of discriminator/generator losses and policy entropy over rounds, thereby confirming that the observed feature consistency and efficiency gains are attainable in practice. revision: yes

  3. Referee: [Table 2] Table 2 or main results table: Reported improvements over federated baselines lack statistical significance tests, standard deviations across multiple runs, or hyperparameter sensitivity analysis; this undermines the central claim that FedRio reliably outperforms under stronger privacy constraints.

    Authors: We agree that the absence of these statistical elements weakens the reliability claims. In the revised results section and Table 2, we will report all main metrics as means with standard deviations computed over five independent runs using different random seeds. We will add p-values from paired statistical tests (e.g., Wilcoxon signed-rank test) comparing FedRio against each federated baseline. Additionally, we will include a hyperparameter sensitivity study for critical parameters including the adversarial loss coefficient, contrastive temperature, and RL learning rate, showing that performance advantages persist across a range of values. These changes will substantiate the robustness under the stronger privacy constraints. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical framework validated on benchmarks

full rationale

The paper proposes a multi-component federated learning system (adaptive GNN message-passing, GAN knowledge extraction, adversarial contrastive learning, RL parameter control) and supports its claims solely through experimental comparisons on two real-world social bot detection benchmarks. No equations, derivations, or first-principles results are presented that reduce by construction to fitted inputs, self-definitions, or self-citation chains. Performance assertions (accuracy, efficiency, consistency) are measured against external baselines and centralized results; they do not rename or re-derive the method's own components. The work is self-contained against external benchmarks with no load-bearing self-citations or ansatz smuggling.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework relies on standard domain assumptions from federated learning and graph neural networks rather than new unproven axioms or invented physical entities; no free parameters are explicitly fitted to produce the central performance claim.

axioms (2)
  • domain assumption Clients are honest and aggregation at the server preserves privacy.
    Implicit in any federated learning setup for social bot detection.
  • domain assumption Graph neural networks with adaptive message passing can serve as a suitable backbone for each client's local social graph.
    Stated as the base architecture for each client.

pith-pipeline@v0.9.0 · 5556 in / 1502 out tokens · 83021 ms · 2026-05-10T15:11:53.440071+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

64 extracted references · 8 canonical work pages · 1 internal anchor

  1. [1]

    Detecting spam in a twitter network,

    S. Yardi, D. Romero, G. Schoenebecket al., “Detecting spam in a twitter network,”First monday, 2010

    2010

  2. [2]

    Dissecting a social botnet: Growth, content and influence in twitter,

    N. Abokhodair, D. Yoo, and D. W. McDonald, “Dissecting a social botnet: Growth, content and influence in twitter,” inCSCW, 2015, pp. 839–851

    2015

  3. [3]

    Perils and challenges of social media and election manipulation analysis: The 2018 us midterms,

    A. Deb, L. Luceri, A. Badaway, and E. Ferrara, “Perils and challenges of social media and election manipulation analysis: The 2018 us midterms,” inWWW, 2019, pp. 237–247

    2018

  4. [4]

    Characterizing social media manipulation in the 2020 us presidential election,

    E. Ferrara, H. Chang, E. Chen, G. Muric, and J. Patel, “Characterizing social media manipulation in the 2020 us presidential election,”First Monday, 2020

    2020

  5. [5]

    A decade of social bot detection,

    S. Cresci, “A decade of social bot detection,”Commun. ACM, vol. 63, no. 10, pp. 72–83, 2020

    2020

  6. [6]

    Online human-bot interactions: Detection, estimation, and characterization,

    O. Varol, E. Ferrara, C. Davis, F. Menczer, and A. Flammini, “Online human-bot interactions: Detection, estimation, and characterization,” in ICWSM, vol. 11, no. 1, 2017, pp. 280–289

    2017

  7. [7]

    The isis twitter census: Defining and describing the population of isis supporters on twitter,

    J. M. Berger and J. Morgan, “The isis twitter census: Defining and describing the population of isis supporters on twitter,” 2015

    2015

  8. [8]

    Predicting online extremism, content adopters, and interaction reciprocity,

    E. Ferrara, W.-Q. Wang, O. Varol, A. Flammini, and A. Galstyan, “Predicting online extremism, content adopters, and interaction reciprocity,” inICSI. Springer, 2016, pp. 22–39

    2016

  9. [9]

    Real-time detection of traffic from twitter stream analysis,

    E. D’Andrea, P. Ducange, B. Lazzerini, and F. Marcelloni, “Real-time detection of traffic from twitter stream analysis,”T-ITS, vol. 16, no. 4, pp. 2269–2283, 2015

    2015

  10. [10]

    Scalable and generalizable social bot detection through data selection,

    K.-C. Yang, O. Varol, P.-M. Hui, and F. Menczer, “Scalable and generalizable social bot detection through data selection,” inAAAI, vol. 34, no. 01, 2020, pp. 1096–1103

    2020

  11. [11]

    Twitter bot detection using bidirectional long short-term memory neural networks and word embeddings,

    F. Wei and U. T. Nguyen, “Twitter bot detection using bidirectional long short-term memory neural networks and word embeddings,” inTPS-ISA. IEEE, 2019, pp. 101–109

    2019

  12. [12]

    Satar: A self-supervised approach to twitter account representation learning and its application in bot detection,

    S. Feng, H. Wan, N. Wang, J. Li, and M. Luo, “Satar: A self-supervised approach to twitter account representation learning and its application in bot detection,” inCIKM, 2021, pp. 3808–3817

    2021

  13. [13]

    Multi-attributed heterogeneous graph convolutional network for bot detection,

    J. Zhao, X. Liu, Q. Yan, B. Li, M. Shao, and H. Peng, “Multi-attributed heterogeneous graph convolutional network for bot detection,”IS, vol. 537, pp. 380–393, 2020

    2020

  14. [14]

    Botmoe: Twitter bot detection with community-aware mixtures of modal-specific experts,

    Y . Liu, Z. Tan, H. Wang, S. Feng, Q. Zheng, and M. Luo, “Botmoe: Twitter bot detection with community-aware mixtures of modal-specific experts,” inSIGIR. ACM, 2023, pp. 485–495

    2023

  15. [15]

    Robctrl: Attacking gnn-based social bot detectors via reinforced manipulation of bots control interaction,

    Y . Yang, X. Zeng, Q. Wu, H. Peng, Y . Xia, H. Liu, B. Chong, and P. S. Yu, “Robctrl: Attacking gnn-based social bot detectors via reinforced manipulation of bots control interaction,”arXiv preprint arXiv:2510.16035, 2025

    work page arXiv 2025

  16. [16]

    Certainly bot or not? trustworthy social bot detection via robust multi-modal neural processes,

    Q. Wu, Y . Yang, H. Peng, B. He, Y . Xia, Y . Liaoet al., “Certainly bot or not? trustworthy social bot detection via robust multi-modal neural processes,”arXiv preprint arXiv:2503.09626, 2025

    work page arXiv 2025

  17. [17]

    Rosgas: Adaptive social bot detection with reinforced self-supervised gnn architecture search,

    Y . Yang, R. Yang, Y . Li, K. Cui, Z. Yang, Y . Wang, J. Xu, and H. Xie, “Rosgas: Adaptive social bot detection with reinforced self-supervised gnn architecture search,”TWEB, 2022

    2022

  18. [18]

    Sebot: Structural entropy guided multi-view contrastive learning for social bot detection,

    Y . Yang, Q. Wu, B. He, H. Peng, R. Yang, Z. Hao, and Y . Liao, “Sebot: Structural entropy guided multi-view contrastive learning for social bot detection,” inProceedings of the 30th ACM SIGKDD conference on knowledge discovery and data mining, 2024, pp. 3841–3852

    2024

  19. [19]

    Dynamicity-aware social bot detection with dynamic graph transformers,

    B. He, Y . Yang, Q. Wu, H. Liu, R. Yang, H. Peng, X. Wang, Y . Liao, and P. Zhou, “Dynamicity-aware social bot detection with dynamic graph transformers,” inProceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, 2024, pp. 5844–5852. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 16

    2024

  20. [20]

    Data-free knowledge distillation for heterogeneous federated learning,

    Z. Zhu, J. Hong, and J. Zhou, “Data-free knowledge distillation for heterogeneous federated learning,” inICML. PMLR, 2021, pp. 12 878– 12 889

    2021

  21. [21]

    Fedgan: Federated generative adversarial networks for distributed data,

    M. Rasouli, T. Sun, and R. Rajagopal, “Fedgan: Federated generative adversarial networks for distributed data,”arXiv, 2020

    2020

  22. [22]

    Fine-tuning global model via data-free knowledge distillation for non-iid federated learning,

    L. Zhang, L. Shen, L. Ding, D. Tao, and L.-Y . Duan, “Fine-tuning global model via data-free knowledge distillation for non-iid federated learning,” inCVPR, 2022, pp. 10 174–10 183

    2022

  23. [23]

    Feddtg: Federated data-free knowledge distillation via three- player generative adversarial networks,

    Z. Zhang, “Feddtg: Federated data-free knowledge distillation via three- player generative adversarial networks,”arXiv, 2022

    2022

  24. [24]

    MH-pFLID: Model heterogeneous personalized federated learning via injection and distillation for medical data analysis,

    L. Xie, M. Lin, S. Liu, C. Xu, Y . Gong, Q. Shen, and Z. Wu, “MH-pFLID: Model heterogeneous personalized federated learning via injection and distillation for medical data analysis,” inICML, 2024

    2024

  25. [25]

    Dual3d-fed: Dual distillation for 3d continual federated learning with vision–language models,

    K. Roy, M. Harandi, and P. Moghadam, “Dual3d-fed: Dual distillation for 3d continual federated learning with vision–language models,”IEEE Access, vol. 13, pp. 204 072–204 086, 2025

    2025

  26. [26]

    Assessing the effects of social familiarity and stance similarity in interaction dynamics,

    K. Dey, R. Shrivastava, S. Kaushik, and V . Mathur, “Assessing the effects of social familiarity and stance similarity in interaction dynamics,” in JCN. Springer, 2018, pp. 843–855

    2018

  27. [27]

    Assessing topical homophily on twitter,

    K. Dey, R. Shrivastava, S. Kaushik, and K. Garg, “Assessing topical homophily on twitter,” inJCN. Springer, 2019, pp. 367–376

    2019

  28. [28]

    Anomalous: A joint modeling approach for anomaly detection on attributed networks

    Z. Peng, M. Luo, J. Li, H. Liu, Q. Zhenget al., “Anomalous: A joint modeling approach for anomaly detection on attributed networks.” in IJCAI, 2018, pp. 3513–3519

    2018

  29. [29]

    On the evolution of user interaction in facebook,

    B. Viswanath, A. Mislove, M. Cha, and K. P. Gummadi, “On the evolution of user interaction in facebook,” inWOSN, 2009, pp. 37–42

    2009

  30. [30]

    Detect me if you can: Spam bot detection using inductive representation learning,

    S. Ali Alhosseini, R. Bin Tareaf, P. Najafi, and C. Meinel, “Detect me if you can: Spam bot detection using inductive representation learning,” inWWW, 2019, pp. 148–153

    2019

  31. [31]

    Botrgcn: Twitter bot detection with relational graph convolutional networks,

    S. Feng, H. Wan, N. Wang, and M. Luo, “Botrgcn: Twitter bot detection with relational graph convolutional networks,” inASONAM, 2021, pp. 236–239

    2021

  32. [32]

    Heterogeneity-aware twitter bot detection with relational graph transformers,

    S. Feng, Z. Tan, R. Li, and M. Luo, “Heterogeneity-aware twitter bot detection with relational graph transformers,” inAAAI, vol. 36, no. 4, 2022, pp. 3977–3985

    2022

  33. [33]

    Social bots detection via fusing bert and graph convolutional networks,

    Q. Guo, H. Xie, Y . Li, W. Ma, and C. Zhang, “Social bots detection via fusing bert and graph convolutional networks,”Symmetry, vol. 14, no. 1, p. 30, 2021

    2021

  34. [34]

    BIC: twitter bot detection with text-graph interaction and semantic consistency,

    Z. Lei, H. Wan, W. Zhang, S. Feng, Z. Chen, J. Li, Q. Zheng, and M. Luo, “BIC: twitter bot detection with text-graph interaction and semantic consistency,” inACL (1). Association for Computational Linguistics, 2023, pp. 10 326–10 340

    2023

  35. [35]

    ETS-MM: A multi-modal social bot detection model based on enhanced textual semantic representation,

    W. Li, J. Deng, J. You, Y . He, Y . Zhuang, and F. Ren, “ETS-MM: A multi-modal social bot detection model based on enhanced textual semantic representation,” inProceedings of the ACM Web Conference 2025, 2025

    2025

  36. [36]

    Lmbot: distilling graph knowledge into language model for graph-less deployment in twitter bot detection,

    Z. Cai, Z. Tan, Z. Lei, Z. Zhu, H. Wang, Q. Zheng, and M. Luo, “Lmbot: distilling graph knowledge into language model for graph-less deployment in twitter bot detection,” inProceedings of the 17th ACM international conference on web search and data mining, 2024, pp. 57–66

    2024

  37. [37]

    Bottrans: A multi-source graph domain adaptation approach for social bot detection,

    B. Shi, Y . Wang, F. Guo, J. Shao, H. Shen, and X. Cheng, “Bottrans: A multi-source graph domain adaptation approach for social bot detection,” inJoint European Conference on Machine Learning and Knowledge Discovery in Databases, 2025, pp. 228–243

    2025

  38. [38]

    Cacl: Community-aware heterogeneous graph contrastive learning for social media bot detection,

    S. Chen, S. Feng, L. Songsong, C.-C. Zong, J. Li, and P. Li, “Cacl: Community-aware heterogeneous graph contrastive learning for social media bot detection,” inFindings of the Association for Computational Linguistics: ACL 2024, 2024, pp. 10 349–10 360

    2024

  39. [39]

    Model compression,

    C. Bucilu ˇa, R. Caruana, and A. Niculescu-Mizil, “Model compression,” inSIGKDD, 2006, pp. 535–541

    2006

  40. [40]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, J. Deanet al., “Distilling the knowledge in a neural network,”arXiv, vol. 2, no. 7, 2015

    2015

  41. [41]

    Federated knowledge distillation,

    H. Seo, J. Park, S. Oh, M. Bennis, and S.-L. Kim, “Federated knowledge distillation,”arXiv, 2020

    2020

  42. [42]

    Ensemble distillation for robust model fusion in federated learning,

    T. Lin, L. Kong, S. U. Stich, and M. Jaggi, “Ensemble distillation for robust model fusion in federated learning,”Advances in Neural Information Processing Systems, vol. 33, pp. 2351–2363, 2020

    2020

  43. [43]

    Fedack: Federated adversarial contrastive knowledge distillation for cross-lingual and cross-model social bot detection,

    Y . Yang, R. Yang, H. Peng, Y . Li, T. Li, Y . Liao, and P. Zhou, “Fedack: Federated adversarial contrastive knowledge distillation for cross-lingual and cross-model social bot detection,” inWWW, 2023

    2023

  44. [44]

    Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,

    A. Fallah, A. Mokhtari, and A. Ozdaglar, “Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,” Advances in Neural Information Processing Systems, vol. 33, pp. 3557– 3568, 2020

    2020

  45. [45]

    Personalized federated learning with moreau envelopes,

    C. T Dinh, N. Tran, and J. Nguyen, “Personalized federated learning with moreau envelopes,”Advances in Neural Information Processing Systems, vol. 33, pp. 21 394–21 405, 2020

    2020

  46. [46]

    Federated evaluation of on-device personalization,

    K. Wang, R. Mathews, C. Kiddon, H. Eichner, F. Beaufays, and D. Ramage, “Federated evaluation of on-device personalization,”arXiv preprint arXiv:1910.10252, 2019

    work page arXiv 1910

  47. [47]

    Federated Learning with Personalization Layers

    M. G. Arivazhagan, V . Aggarwal, A. K. Singh, and S. Choud- hary, “Federated learning with personalization layers,”arXiv preprint arXiv:1912.00818, 2019

    work page internal anchor Pith review arXiv 1912

  48. [48]

    Salvaging federated learning by local adaptation,

    T. Yu, E. Bagdasaryan, and V . Shmatikov, “Salvaging federated learning by local adaptation,”arXiv preprint arXiv:2002.04758, 2020

    work page arXiv 2002

  49. [49]

    Three approaches for personalization with applications to federated learning

    Y . Mansour, M. Mohri, J. Ro, and A. T. Suresh, “Three approaches for personalization with applications to federated learning,”arXiv preprint arXiv:2002.10619, 2020

    work page arXiv 2002

  50. [50]

    Adaptive personalized fed- erated learning.arXiv preprint arXiv:2003.13461,

    Y . Deng, M. M. Kamani, and M. Mahdavi, “Adaptive personalized federated learning,”arXiv preprint arXiv:2003.13461, 2020

    work page arXiv 2003

  51. [51]

    Personalized fed- erated learning with first order model optimization.arXiv preprint arXiv:2012.08565,

    M. Zhang, K. Sapra, S. Fidler, S. Yeung, and J. M. Alvarez, “Personalized federated learning with first order model optimization,”arXiv preprint arXiv:2012.08565, 2020

    work page arXiv 2012

  52. [52]

    Per- sonalized cross-silo federated learning on non-iid data,

    Y . Huang, L. Chu, Z. Zhou, L. Wang, J. Liu, J. Pei, and Y . Zhang, “Per- sonalized cross-silo federated learning on non-iid data,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 9, 2021, pp. 7865–7873

    2021

  53. [53]

    An efficient frame- work for clustered federated learning,

    A. Ghosh, J. Chung, D. Yin, and K. Ramchandran, “An efficient frame- work for clustered federated learning,”Advances in Neural Information Processing Systems, vol. 33, pp. 19 586–19 597, 2020

    2020

  54. [54]

    Parameterized knowledge transfer for personalized federated learning,

    J. Zhang, S. Guo, X. Ma, H. Wang, W. Xu, and F. Wu, “Parameterized knowledge transfer for personalized federated learning,”Advances in Neural Information Processing Systems, vol. 34, pp. 10 092–10 104, 2021

    2021

  55. [55]

    Federated multi-task learning,

    V . Smith, C.-K. Chiang, M. Sanjabi, and A. S. Talwalkar, “Federated multi-task learning,”Advances in neural information processing systems, vol. 30, 2017

    2017

  56. [56]

    Fedssp: federated graph learning with spectral knowledge and personalized preference,

    Z. Tan, G. Wan, W. Huang, and M. Ye, “Fedssp: federated graph learning with spectral knowledge and personalized preference,”Advances in Neural Information Processing Systems, vol. 37, pp. 34 561–34 581, 2024

    2024

  57. [57]

    Adpfedgnn: Adaptive decoupling personalized federated graph neural network,

    Z. Guan, Y . Li, J. Du, R. Tang, and X. Meng, “Adpfedgnn: Adaptive decoupling personalized federated graph neural network,” inProceed- ings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, 2025, pp. 5253–5261

    2025

  58. [58]

    Arming the public with artificial intelligence to counter social bots,

    K.-C. Yang, O. Varol, C. A. Davis, E. Ferrara, A. Flammini, and F. Menczer, “Arming the public with artificial intelligence to counter social bots,”Comput. Hum. Behav., vol. 1, no. 1, pp. 48–61, 2019

    2019

  59. [59]

    Twibot-20: A comprehensive twitter bot detection benchmark,

    S. Feng, H. Wan, N. Wang, J. Li, and M. Luo, “Twibot-20: A comprehensive twitter bot detection benchmark,” inCIKM, 2021, pp. 4485–4494

    2021

  60. [60]

    Federated learning on non-iid data silos: An experimental study,

    Q. Li, Y . Diao, Q. Chen, and B. He, “Federated learning on non-iid data silos: An experimental study,”arXiv, 2021

    2021

  61. [61]

    Communication-efficient learning of deep networks from decentralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inAISTATS. PMLR, 2017, pp. 1273–1282

    2017

  62. [62]

    Federated optimization in heterogeneous networks,

    T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V . Smith, “Federated optimization in heterogeneous networks,”PMLS, vol. 2, pp. 429–450, 2020

    2020

  63. [63]

    Fed-ensemble: Im- proving generalization through model ensembling in federated learning,

    N. Shi, F. Lai, R. A. Kontar, and M. Chowdhury, “Fed-ensemble: Im- proving generalization through model ensembling in federated learning,” arXiv, 2021

    2021

  64. [64]

    Twibot- 22: Towards graph-based twitter bot detection,

    S. Feng, Z. Tan, H. Wan, N. Wang, Z. Chen, B. Zhang, Q. Zheng, W. Zhang, Z. Lei, S. Yang, X. Feng, Q. Zhang, H. Wang, Y . Liu, Y . Bai, H. Wang, Z. Cai, Y . Wang, L. Zheng, Z. Ma, J. Li, and M. Luo, “Twibot- 22: Towards graph-based twitter bot detection,” inNeurIPS, 2022. Yingguang Yangreceived the Ph.D. degree in the School of Cyber Science at University...

    2022