pith. sign in

arxiv: 2605.02935 · v1 · submitted 2026-04-30 · 💻 cs.LG · cs.AI

DeRelayL: Sustainable Decentralized Relay Learning

Haihan Duan , Tengfei Ma , Yuyang Qin , Runhao Zeng , Wei Cai , Victor C. M. Leung , Xiping Hu This is my paper

Pith reviewed 2026-05-09 20:21 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords decentralized learningrelay learningincentive mechanismscollaborative model trainingpermissionless participationsustainable AI systemslarge-scale modelsalternative to federated learning
0
0 comments X

The pith

DeRelayL lets permissionless users train and share large models by relaying contributions in a decentralized system.

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

Large-scale model training currently requires resources only available to big institutions, excluding everyday users who generate much of the data. The paper introduces decentralized relay learning as a way for anyone to join without permission, pass an evolving model along in sequence, add their training contribution, and receive the final shared model. Incentive mechanisms are built in to encourage continued participation and keep the system running without a central organizer. Architecture details, workflow steps, theoretical proofs, and simulations are provided to support that the approach maintains effectiveness. If accurate, this would let ordinary participants own and benefit from models previously controlled by a few entities.

Core claim

Decentralized relay learning is a sustainable system in which permissionless participants contribute to model training in a relay-like manner and share the resulting model, with incentive mechanisms ensuring ongoing participation as shown through the described architecture, workflow, theoretical analysis, and numerical simulations.

What carries the argument

The DeRelayL architecture in which an initial model is passed sequentially among participants, each performs local training updates before relaying it onward, and incentives are distributed to sustain the chain.

If this is right

  • Common users gain the ability to contribute to and own portions of large models without central approval or high personal costs.
  • The relay process distributes computational load across many devices while preserving model quality through sequential updates.
  • Incentive designs address the free-rider problem that often undermines decentralized collaborative training.
  • The system provides an alternative to federated learning by emphasizing sequential passing and explicit sustainability mechanisms over group aggregation.

Where Pith is reading between the lines

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

  • This relay structure could allow models to improve incrementally from distributed user data sources without requiring all participants to be online simultaneously.
  • It might scale to domains like mobile sensor data or personal health records where users want to pool models privately.
  • Real deployments would need to test whether the theoretical incentives hold when participants have varying device capabilities or attempt to minimize their effort.

Load-bearing premise

Incentive mechanisms can be designed to reliably ensure ongoing participation and sustainability in a fully permissionless decentralized setting.

What would settle it

A controlled simulation or small-scale deployment in which participation rates fall sharply or model performance stagnates despite the proposed incentives, contradicting the numerical results.

Figures

Figures reproduced from arXiv: 2605.02935 by Haihan Duan, Runhao Zeng, Tengfei Ma, Victor C. M. Leung, Wei Cai, Xiping Hu, Yuyang Qin.

Figure 1
Figure 1. Figure 1: A comparative diagram between federated learning and relay learning. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: System architecture and workflow of sustainable decentralized relay learning (DeRelayL). [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The observed trend indicates that the growth rate of [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Model version distribution over rounds [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
read the original abstract

In the era of big data, large-scale machine learning models have revolutionized various fields, driving significant advancements. However, large-scale model training demands high financial and computational resources, which are only affordable by a few technological giants and well-funded institutions. In this case, common users like mobile users, the real creators of valuable data, are often excluded from fully benefiting due to the barriers, while the current methods for accessing large-scale models either limit user ownership or lack sustainability. This growing gap highlights the urgent need for a collaborative model training approach, allowing common users to train and share models. However, existing collaborative model training paradigms, especially federated learning (FL), primarily focus on data privacy and group-based model aggregation. To this end, this paper intends to address this issue by proposing a novel training paradigm named decentralized relay learning (DeRelayL), a sustainable learning system where permissionless participants can contribute to model training in a relay-like manner and share the model. In detail, this paper presents the architecture and workflow of DeRelayL, designs incentive mechanisms to ensure sustainability, and conducts theoretical analysis and numerical simulations to demonstrate its effectiveness.

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 DeRelayL, a novel decentralized relay learning paradigm for sustainable collaborative training of large-scale ML models. Permissionless participants contribute to model training in a relay-like manner and share the resulting model. The manuscript presents the system architecture and workflow, designs incentive mechanisms to promote ongoing participation, and supports the claims with theoretical analysis plus numerical simulations demonstrating effectiveness.

Significance. If the incentive mechanisms and theoretical results hold, the work could meaningfully advance decentralized ML by enabling broader, sustainable participation beyond centralized or federated approaches, directly addressing resource barriers for common users. The explicit focus on sustainability via incentives, combined with architecture, theory, and simulations, represents a constructive contribution to the field.

major comments (3)
  1. [§4] §4 (Incentive Mechanism Design): the sustainability claim rests on the designed incentives creating stable participation equilibria, yet no game-theoretic analysis, Nash equilibrium derivation, or robustness proof against free-riding, Sybil attacks, or heterogeneous churn is provided. This is load-bearing for the central claim that the system remains sustainable in a fully permissionless setting without central coordination.
  2. [§5] §5 (Theoretical Analysis): the analysis is stated to demonstrate effectiveness, but the manuscript does not specify the key assumptions (e.g., participant utility functions, enforcement mechanisms, or relay dynamics) or show that the derived predictions are non-circular or independent of fitted parameters. Without these, it is impossible to confirm whether the theory supports generalization beyond the simulated scenarios.
  3. [§6] §6 (Numerical Simulations): simulations are used to validate effectiveness, but the manuscript provides no details on baselines, metrics, error bars, number of runs, or exclusion criteria for participant behavior. This prevents assessment of whether the results robustly support the sustainability and relay-learning claims under realistic conditions.
minor comments (2)
  1. The abstract and introduction should more explicitly contrast DeRelayL with federated learning and existing decentralized training methods, including specific differences in model sharing and relay workflow.
  2. Notation for relay steps, model versions, and incentive parameters should be defined consistently in a dedicated notation table or early section to improve readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which identify key areas where additional rigor will strengthen the manuscript. We address each major comment below, indicating the revisions we will make to the next version.

read point-by-point responses

  1. Referee: [§4] §4 (Incentive Mechanism Design): the sustainability claim rests on the designed incentives creating stable participation equilibria, yet no game-theoretic analysis, Nash equilibrium derivation, or robustness proof against free-riding, Sybil attacks, or heterogeneous churn is provided. This is load-bearing for the central claim that the system remains sustainable in a fully permissionless setting without central coordination.

    Authors: We agree that the sustainability claims would be substantially strengthened by explicit game-theoretic analysis. The current incentive mechanisms are motivated by economic alignment and supported by simulation outcomes, but the manuscript indeed omits formal Nash equilibrium derivations and robustness proofs. In the revised manuscript we will add a new subsection to §4 that derives the Nash equilibria for the proposed incentive structure and analyzes stability against free-riding, Sybil attacks, and heterogeneous churn. revision: yes

  2. Referee: [§5] §5 (Theoretical Analysis): the analysis is stated to demonstrate effectiveness, but the manuscript does not specify the key assumptions (e.g., participant utility functions, enforcement mechanisms, or relay dynamics) or show that the derived predictions are non-circular or independent of fitted parameters. Without these, it is impossible to confirm whether the theory supports generalization beyond the simulated scenarios.

    Authors: We acknowledge that the theoretical section requires clearer exposition of assumptions and verification of non-circularity. The analysis relies on standard participant utility models and relay dynamics, yet these were not stated explicitly. In the revision we will open §5 with a dedicated assumptions paragraph that defines participant utility functions, enforcement mechanisms, and relay dynamics, and we will demonstrate that the derived predictions follow directly from the model without dependence on simulation-fitted parameters. revision: yes

  3. Referee: [§6] §6 (Numerical Simulations): simulations are used to validate effectiveness, but the manuscript provides no details on baselines, metrics, error bars, number of runs, or exclusion criteria for participant behavior. This prevents assessment of whether the results robustly support the sustainability and relay-learning claims under realistic conditions.

    Authors: We agree that the simulation section lacks sufficient methodological detail for reproducibility and robustness evaluation. The experiments compare DeRelayL against centralized and federated baselines using participation rate, model accuracy, and resource metrics, but these elements were not documented. In the revised §6 we will add explicit descriptions of the baselines, evaluation metrics, error bars computed over 10 independent runs, the total number of simulation runs, and the modeling choices for participant behavior including churn and heterogeneity. revision: yes

Circularity Check

0 steps flagged

No circularity: DeRelayL claims rest on design and simulation, not self-referential derivation

full rationale

The paper introduces DeRelayL as a new paradigm with architecture, workflow, incentive mechanisms, theoretical analysis, and simulations. No equations, derivations, or load-bearing steps appear in the abstract or description that reduce by construction to fitted parameters, self-definitions, or self-citations. The sustainability argument is presented as a designed outcome validated empirically, which is independent content rather than tautological. This is the expected non-finding for a system-proposal paper whose central claims do not invoke uniqueness theorems or rename known results via internal loops.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no specific free parameters, axioms, or invented entities can be extracted. The proposal implicitly assumes that incentive mechanisms can be designed to sustain permissionless relay participation without central authority.

pith-pipeline@v0.9.0 · 5512 in / 1263 out tokens · 36909 ms · 2026-05-09T20:21:54.300086+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

60 extracted references · 60 canonical work pages · 2 internal anchors

  1. [1]

    A survey on evaluation of large language models,

    Y . Chang, X. Wang, J. Wang, Y . Wu, L. Yang, K. Zhu, H. Chen, X. Yi, C. Wang, Y . Wanget al., “A survey on evaluation of large language models,”ACM Transactions on Intelligent Systems and Technology, vol. 15, no. 3, pp. 1–45, 2024

    work page 2024

  2. [2]

    Scaling Laws for Neural Language Models

    J. Kaplan, S. McCandlish, T. Henighan, T. B. Brown, B. Chess, R. Child, S. Gray, A. Radford, J. Wu, and D. Amodei, “Scaling laws for neural language models,”arXiv preprint arXiv:2001.08361, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2001

  3. [3]

    When scaling meets llm finetuning: The effect of data, model and finetuning method,

    B. Zhang, Z. Liu, C. Cherry, and O. Firat, “When scaling meets llm finetuning: The effect of data, model and finetuning method,” inThe Twelfth International Conference on Learning Representations, 2024

    work page 2024

  4. [4]

    Incentive mechanism design toward a win–win situation for generative art trainers and artists,

    H. Duan, A. El Saddik, and W. Cai, “Incentive mechanism design toward a win–win situation for generative art trainers and artists,”IEEE Transactions on Computational Social Systems, 2024

    work page 2024

  5. [5]

    Free open source communities sustainability: Does it make a difference in software quality?

    A. Alami, R. Pardo, and J. Lin ˚aker, “Free open source communities sustainability: Does it make a difference in software quality?”Empirical Software Engineering, vol. 29, no. 5, p. 114, 2024

    work page 2024

  6. [6]

    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,” inArtificial intelligence and statistics. PMLR, 2017, pp. 1273– 1282

    work page 2017

  7. [7]

    Optimal power control for over-the-air federated learning with gradient compression,

    M. Ruan, Y . Li, W. Zhang, L. Song, and W. Xu, “Optimal power control for over-the-air federated learning with gradient compression,” in2024 IEEE 30th International Conference on Parallel and Distributed Systems (ICPADS). IEEE, 2024, pp. 326–333

    work page 2024

  8. [8]

    The non-iid data quagmire of decentralized machine learning,

    K. Hsieh, A. Phanishayee, O. Mutlu, and P. Gibbons, “The non-iid data quagmire of decentralized machine learning,” inInternational Conference on Machine Learning. PMLR, 2020, pp. 4387–4398

    work page 2020

  9. [9]

    Scaffold: Stochastic controlled averaging for federated learn- ing,

    S. P. Karimireddy, S. Kale, M. Mohri, S. Reddi, S. Stich, and A. T. Suresh, “Scaffold: Stochastic controlled averaging for federated learn- ing,” inInternational conference on machine learning. PMLR, 2020, pp. 5132–5143

    work page 2020

  10. [10]

    Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges,

    E. T. M. Beltr ´an, M. Q. P ´erez, P. M. S. S ´anchez, S. L. Bernal, G. Bovet, M. G. P ´erez, G. M. P ´erez, and A. H. Celdr ´an, “Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges,”IEEE Communications Surveys & Tutorials, 2023

    work page 2023

  11. [11]

    Blockchain-enabled federated learning: A survey,

    Y . Qu, M. P. Uddin, C. Gan, Y . Xiang, L. Gao, and J. Yearwood, “Blockchain-enabled federated learning: A survey,”ACM Computing Surveys, vol. 55, no. 4, pp. 1–35, 2022

    work page 2022

  12. [12]

    Cola: Decentralized linear learning,

    L. He, A. Bian, and M. Jaggi, “Cola: Decentralized linear learning,” Advances in Neural Information Processing Systems, vol. 31, 2018

    work page 2018

  13. [13]

    Learning to collaborate in decentralized learning of personalized models,

    S. Li, T. Zhou, X. Tian, and D. Tao, “Learning to collaborate in decentralized learning of personalized models,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 9766–9775

    work page 2022

  14. [14]

    Blockchained on-device federated learning,

    H. Kim, J. Park, M. Bennis, and S.-L. Kim, “Blockchained on-device federated learning,”IEEE Communications Letters, vol. 24, no. 6, pp. 1279–1283, 2019

    work page 2019

  15. [15]

    A blockchain- based decentralized federated learning framework with committee con- sensus,

    Y . Li, C. Chen, N. Liu, H. Huang, Z. Zheng, and Q. Yan, “A blockchain- based decentralized federated learning framework with committee con- sensus,”IEEE Network, vol. 35, no. 1, pp. 234–241, 2020

    work page 2020

  16. [16]

    Proof of federated learning: A novel energy-recycling consensus algorithm,

    X. Qu, S. Wang, Q. Hu, and X. Cheng, “Proof of federated learning: A novel energy-recycling consensus algorithm,”IEEE Transactions on Parallel and Distributed Systems, vol. 32, no. 8, pp. 2074–2085, 2021

    work page 2074

  17. [17]

    Baffle: Blockchain based aggregator free federated learning,

    P. Ramanan and K. Nakayama, “Baffle: Blockchain based aggregator free federated learning,” in2020 IEEE international conference on blockchain (Blockchain). IEEE, 2020, pp. 72–81

    work page 2020

  18. [18]

    Blockchain-based node-aware dynamic weighting methods for improving federated learning performance,

    Y . J. Kim and C. S. Hong, “Blockchain-based node-aware dynamic weighting methods for improving federated learning performance,” in 2019 20th Asia-pacific network operations and management symposium (APNOMS). IEEE, 2019, pp. 1–4. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 16

    work page 2019

  19. [19]

    Hybrid blockchain-based resource trading system for federated learning in edge computing,

    S. Fan, H. Zhang, Y . Zeng, and W. Cai, “Hybrid blockchain-based resource trading system for federated learning in edge computing,”IEEE Internet of Things Journal, vol. 8, no. 4, pp. 2252–2264, 2020

    work page 2020

  20. [20]

    Blockchain assisted decentralized federated learning (blade-fl): Performance analysis and resource allocation,

    J. Li, Y . Shao, K. Wei, M. Ding, C. Ma, L. Shi, Z. Han, and H. V . Poor, “Blockchain assisted decentralized federated learning (blade-fl): Performance analysis and resource allocation,”IEEE Transactions on Parallel and Distributed Systems, vol. 33, no. 10, pp. 2401–2415, 2021

    work page 2021

  21. [21]

    Blockdfl: A blockchain- based fully decentralized peer-to-peer federated learning framework,

    Z. Qin, X. Yan, M. Zhou, and S. Deng, “Blockdfl: A blockchain- based fully decentralized peer-to-peer federated learning framework,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 2914– 2925

    work page 2024

  22. [22]

    Towards decentralised learning analytics (positioning paper),

    A. Ekuban and J. Domingue, “Towards decentralised learning analytics (positioning paper),” inCompanion Proceedings of the ACM Web Conference 2023, 2023, pp. 1435–1438

    work page 2023

  23. [23]

    Incentive mechanism for horizontal fed- erated learning based on reputation and reverse auction,

    J. Zhang, Y . Wu, and R. Pan, “Incentive mechanism for horizontal fed- erated learning based on reputation and reverse auction,” inProceedings of the Web Conference 2021, 2021, pp. 947–956

    work page 2021

  24. [24]

    Distributed deep learning in open collaborations,

    M. Diskin, A. Bukhtiyarov, M. Ryabinin, L. Saulnier, A. Sinitsin, D. Popov, D. V . Pyrkin, M. Kashirin, A. Borzunov, A. Villanova del Moralet al., “Distributed deep learning in open collaborations,”Ad- vances in Neural Information Processing Systems, vol. 34, pp. 7879– 7897, 2021

    work page 2021

  25. [25]

    Towards crowdsourced training of large neural networks using decentralized mixture-of-experts,

    M. Ryabinin and A. Gusev, “Towards crowdsourced training of large neural networks using decentralized mixture-of-experts,”Advances in Neural Information Processing Systems, vol. 33, pp. 3659–3672, 2020

    work page 2020

  26. [26]

    Cerebro: A platform for multi-party cryptographic collaborative learn- ing,

    W. Zheng, R. Deng, W. Chen, R. A. Popa, A. Panda, and I. Stoica, “Cerebro: A platform for multi-party cryptographic collaborative learn- ing,” in30th USENIX Security Symposium (USENIX Security 21), 2021, pp. 2723–2740

    work page 2021

  27. [27]

    Tiny multi-agent drl for twins migration in uav metaverses: A multi-leader multi-follower stackelberg game approach,

    J. Kang, Y . Zhong, M. Xu, J. Nie, J. Wen, H. Du, D. Ye, X. Huang, D. Niyato, and S. Xie, “Tiny multi-agent drl for twins migration in uav metaverses: A multi-leader multi-follower stackelberg game approach,” IEEE Internet of Things Journal, 2024

    work page 2024

  28. [28]

    Gradientcoin: A peer-to-peer decentralized large language models

    Y . Gao, Z. Song, and J. Yin, “Gradientcoin: A peer-to-peer decentralized large language models,”arXiv preprint arXiv:2308.10502, 2023

    work page arXiv 2023

  29. [29]

    Relay learning: a physically secure framework for clinical multi-site deep learning,

    Z.-H. Bo, Y . Guo, J. Lyu, H. Liang, J. He, S. Deng, F. Xu, X. Lou, and Q. Dai, “Relay learning: a physically secure framework for clinical multi-site deep learning,”NPJ Digital Medicine, vol. 6, no. 1, p. 204, 2023

    work page 2023

  30. [30]

    Clustered sampling: Low-variance and improved representativity for clients selection in federated learning,

    Y . Fraboni, R. Vidal, L. Kameni, and M. Lorenzi, “Clustered sampling: Low-variance and improved representativity for clients selection in federated learning,” inInternational Conference on Machine Learning. PMLR, 2021, pp. 3407–3416

    work page 2021

  31. [31]

    Federated learning with matched averaging,

    H. Wang, M. Yurochkin, Y . Sun, D. Papailiopoulos, and Y . Khazaeni, “Federated learning with matched averaging,”International Conference on Learning Representations, 2020

    work page 2020

  32. [32]

    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

    work page 2020

  33. [33]

    Federated optimization in heterogeneous networks,

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

    work page 2020

  34. [34]

    Fedbe: Making bayesian model ensemble applicable to federated learning,

    H.-Y . Chen and W.-L. Chao, “Fedbe: Making bayesian model ensemble applicable to federated learning,”International Conference on Learning Representations, 2021

    work page 2021

  35. [35]

    Page: Equilibrate personalization and generalization in federated learning,

    Q. Chen, Z. Wang, J. Hu, H. Yan, J. Zhou, and X. Lin, “Page: Equilibrate personalization and generalization in federated learning,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 2955–2964

    work page 2024

  36. [36]

    Accelerating the decentralized federated learning via manipulating edges,

    M. Zhou, G. Liu, K. Lu, R. Mao, and H. Liao, “Accelerating the decentralized federated learning via manipulating edges,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 2945–2954

    work page 2024

  37. [37]

    Fededge: Accelerating edge-assisted federated learning,

    K. Wang, Q. He, F. Chen, H. Jin, and Y . Yang, “Fededge: Accelerating edge-assisted federated learning,” inProceedings of the ACM Web Conference 2023, 2023, pp. 2895–2904

    work page 2023

  38. [38]

    An energy-efficient and privacy- aware decomposition framework for edge-assisted federated learning,

    Y . Shi, H. Duan, L. Yang, and W. Cai, “An energy-efficient and privacy- aware decomposition framework for edge-assisted federated learning,” ACM Transactions on Sensor Networks, vol. 18, no. 4, pp. 1–24, 2022

    work page 2022

  39. [39]

    Privacy-preserving and fairness-aware federated learning for critical infrastructure protection and resilience,

    Y . Zhang, R. Sun, L. Shen, G. Bai, M. Xue, M. H. Meng, X. Li, R. Ko, and S. Nepal, “Privacy-preserving and fairness-aware federated learning for critical infrastructure protection and resilience,” inProceedings of the ACM on Web Conference 2024, 2024, pp. 2986–2997

    work page 2024

  40. [40]

    Toward on-device federated learning: A direct acyclic graph-based blockchain approach,

    M. Cao, L. Zhang, and B. Cao, “Toward on-device federated learning: A direct acyclic graph-based blockchain approach,”IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 4, pp. 2028–2042, 2021

    work page 2028

  41. [41]

    Proof-of-contribution-based design for collaborative machine learning on blockchain,

    B. Buyukates, C. He, S. Han, Z. Fang, Y . Zhang, J. Long, A. Farahanchi, and S. Avestimehr, “Proof-of-contribution-based design for collaborative machine learning on blockchain,” in2023 IEEE International Con- ference on Decentralized Applications and Infrastructures (DAPPS). IEEE, 2023, pp. 13–22

    work page 2023

  42. [42]

    Blockchain-based federated learning utilizing zero- knowledge proofs for verifiable training and aggregation,

    E. Ebrahimi, M. Sober, A.-T. Hoang, C. U. Ileri, W. Sanders, and S. Schulte, “Blockchain-based federated learning utilizing zero- knowledge proofs for verifiable training and aggregation,” in2024 IEEE International Conference on Blockchain (Blockchain). IEEE, 2024, pp. 54–63

    work page 2024

  43. [43]

    A blockchain-enabled and transparent evaluation of ml models in the decentralised marketplace,

    H. Yazdaninejad, M. Rajarajan, and M. Krol, “A blockchain-enabled and transparent evaluation of ml models in the decentralised marketplace,” in2024 IEEE International Conference on Blockchain (Blockchain). IEEE, 2024, pp. 458–463

    work page 2024

  44. [44]

    Federated machine learning: Concept and applications,

    Q. Yang, Y . Liu, T. Chen, and Y . Tong, “Federated machine learning: Concept and applications,”ACM Transactions on Intelligent Systems and Technology (TIST), vol. 10, no. 2, pp. 1–19, 2019

    work page 2019

  45. [45]

    A survey on fully homomor- phic encryption: An engineering perspective,

    P. Martins, L. Sousa, and A. Mariano, “A survey on fully homomor- phic encryption: An engineering perspective,”ACM Computing Surveys (CSUR), vol. 50, no. 6, pp. 1–33, 2017

    work page 2017

  46. [46]

    Sharp: A short- word hierarchical accelerator for robust and practical fully homomorphic encryption,

    J. Kim, S. Kim, J. Choi, J. Park, D. Kim, and J. H. Ahn, “Sharp: A short- word hierarchical accelerator for robust and practical fully homomorphic encryption,” inProceedings of the 50th Annual International Symposium on Computer Architecture, 2023, pp. 1–15

    work page 2023

  47. [47]

    Bitcoin: A peer-to-peer electronic cash system,

    S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,”Satoshi Nakamoto, 2008

    work page 2008

  48. [48]

    The double auction market institution: A survey,

    D. Friedman, “The double auction market institution: A survey,” inThe double auction market. Routledge, 2018, pp. 3–26

    work page 2018

  49. [49]

    An efficient auction- based mechanism for mobile data offloading,

    S. Paris, F. Martignon, I. Filippini, and L. Chen, “An efficient auction- based mechanism for mobile data offloading,”IEEE Transactions on Mobile Computing, vol. 14, no. 8, pp. 1573–1586, 2014

    work page 2014

  50. [50]

    A profit-maximizing model marketplace with differentially private federated learning,

    P. Sun, X. Chen, G. Liao, and J. Huang, “A profit-maximizing model marketplace with differentially private federated learning,” inIEEE IN- FOCOM 2022-IEEE Conference on Computer Communications. IEEE, 2022, pp. 1439–1448

    work page 2022

  51. [51]

    De- centralized applications: The blockchain-empowered software system,

    W. Cai, Z. Wang, J. B. Ernst, Z. Hong, C. Feng, and V . C. Leung, “De- centralized applications: The blockchain-empowered software system,” IEEE access, vol. 6, pp. 53 019–53 033, 2018

    work page 2018

  52. [52]

    Analysis of security in blockchain: Case study in 51%-attack detecting,

    C. Ye, G. Li, H. Cai, Y . Gu, and A. Fukuda, “Analysis of security in blockchain: Case study in 51%-attack detecting,” in2018 5th Interna- tional conference on dependable systems and their applications (DSA). IEEE, 2018, pp. 15–24

    work page 2018

  53. [53]

    Blockchain-based decentralized reputation system in e-commerce envi- ronment,

    Z. Zhou, M. Wang, C.-N. Yang, Z. Fu, X. Sun, and Q. J. Wu, “Blockchain-based decentralized reputation system in e-commerce envi- ronment,”Future Generation Computer Systems, vol. 124, pp. 155–167, 2021

    work page 2021

  54. [54]

    Language mod- els are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language mod- els are few-shot learners,”Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

    work page 1901

  55. [55]

    The Llama 3 Herd of Models

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fanet al., “The llama 3 herd of models,”arXiv preprint arXiv:2407.21783, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  56. [56]

    Zkcnn: Zero knowledge proofs for con- volutional neural network predictions and accuracy,

    T. Liu, X. Xie, and Y . Zhang, “Zkcnn: Zero knowledge proofs for con- volutional neural network predictions and accuracy,” inProceedings of the 2021 ACM SIGSAC Conference on Computer and Communications Security, 2021, pp. 2968–2985

    work page 2021

  57. [57]

    Mystique: Efficient conversions for zero-knowledge proofs with applications to machine learning,

    C. Weng, K. Yang, X. Xie, J. Katz, and X. Wang, “Mystique: Efficient conversions for zero-knowledge proofs with applications to machine learning,” in30th USENIX Security Symposium (USENIX Security 21), 2021, pp. 501–518

    work page 2021

  58. [58]

    Data valuation and pricing in internet of things: Survey and vision,

    X. Shi and H. Duan, “Data valuation and pricing in internet of things: Survey and vision,” in2024 IEEE International Conference on Smart Internet of Things (SmartIoT). IEEE, 2024, pp. 547–554. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 17 Haihan Duan(Member, IEEE) received his B.Eng. degree in Computer Science and Technology from East Ch...

    work page 2024

  59. [59]

    Individual Rationality (IR):To achieve IR in the DeRe- layL system, all participants that choose the “Normal” strategy should at least have positive utilities, which means that the incentive provided by the proposed mechanism should lead to U P articipant N ≥0. Therefore, for each participant, there will be some conditions to guarantee thatU P articipant ...

    work page 2020

  60. [60]

    Normal” strategy, which means that the utility of the “Normal

    Incentive Compatibility (IC):To achieve IC in the DeRelayL system, all rational participants will tend to choose the “Normal” strategy, which means that the utility of the “Normal” strategy should be greater than other strategies. Therefore, for each participant, the incentive provided by the proposed mechanism should lead toU P articipant N ≥ U P articip...

    work page 2020