pith. sign in

arxiv: 2606.12246 · v1 · pith:GIPPZ6SKnew · submitted 2026-06-10 · 💻 cs.DC · cs.IR

Efficient and Robust Online Learning to Rank in Decentralized Systems

Marcel Gregoriadis , Martijn de Vos , Sayan Biswas , Anne-Marie Kermarrec , Johan Pouwelse This is my paper

Pith reviewed 2026-06-27 08:14 UTC · model grok-4.3

classification 💻 cs.DC cs.IR
keywords online learning to rankdecentralized learningpoisoning attacksrobust aggregationconvergence guaranteeposition bias correctionclick models
0
0 comments X

The pith

RankGuard aggregates a decentralized ranking model only if it better explains the user's private click history after position bias correction.

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

Decentralized online learning to rank lets users exchange model updates directly without a central server. RankGuard protects against poisoned updates by letting each user test an incoming model on their own past clicks, after correcting for position bias, and accept it only when it fits those clicks better than the current local model. The paper proves that this rule yields convergence of the shared model. Experiments across four benchmarks and three click models show RankGuard resists four poisoning attacks, including adaptive ones, and runs up to 62 times faster than prior defenses. A reader would care because the approach removes the need for a trusted coordinator while keeping malicious updates from degrading local ranking quality.

Core claim

RankGuard is a decentralized OLTR framework in which each node evaluates every incoming model against its private click history after position-bias correction and aggregates the update only when the new model explains the observed clicks better than the current local model; this rule supplies both robustness to poisoning and the first formal convergence guarantee for any decentralized OLTR algorithm.

What carries the argument

The acceptance test that compares how well an incoming model explains the user's position-bias-corrected private click history versus the current local model.

If this is right

  • The algorithm is guaranteed to converge under the stated conditions.
  • It resists four poisoning attacks, including a strong adaptive attack, across four standard benchmarks.
  • It outperforms prior defenses in most tested settings while using up to 62 times less computation.
  • The same acceptance rule works with three different click models.

Where Pith is reading between the lines

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

  • The same local-data test could be applied to other decentralized learning tasks where each participant holds private interaction logs.
  • Stronger or more accurate position-bias correction would directly tighten the defense without changing the aggregation logic.
  • The method reduces reliance on any single trusted server, which may matter for recommendation systems that must operate across distrusting organizations.

Load-bearing premise

A user's private click history, once corrected for position bias, supplies a reliable signal of whether an incoming model will genuinely improve local ranking quality.

What would settle it

A concrete counter-example would be an attack that produces a model passing the acceptance test on every honest node's history yet produces measurably worse ranking quality on held-out queries from those same nodes.

Figures

Figures reproduced from arXiv: 2606.12246 by Anne-Marie Kermarrec, Johan Pouwelse, Marcel Gregoriadis, Martijn de Vos, Sayan Biswas.

Figure 1
Figure 1. Figure 1: Ranking accuracy of honest nodes as users interact [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The workflow of RankGuard, our decentralized framework for robust OLTR. 3.2 RankGuard in a Nutshell The key question when doing OLTR using GL is how a node should weigh a single model received from another node when there is no trusted server that has access to multiple models to run and no way to tell which nodes are malicious. RankGuard answers this by treating each node’s own click history as a private,… view at source ↗
Figure 3
Figure 3. Figure 3: Robustness under the Adapt attack. test set after every full round (100 sessions), reporting the mean and standard deviation across nodes. 6 Experimental Evaluation We now present the experimental evaluation of RankGuard, which answers the following questions: (1) What is the robustness of RankGuard on preserving rank￾ing accuracy compared to defense baselines, for different attacks, click models, and data… view at source ↗
Figure 4
Figure 4. Figure 4: Evolution of 𝛼 for different attacks from honest vs. malicious users [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Robustness of RankGuard under varying shares of malicious nodes in the network. reduces to small noise around the honest mean, which under the already-noisy informational signal acts as mild regularization rather than a genuine attack. Under IPM with the navigational click model, 𝛽 = 0.5 stalls convergence entirely, plateauing at nDCG@10 ≈ 0.25 while all lower attack shares reach ≈ 0.4, indicating the need… view at source ↗
Figure 6
Figure 6. Figure 6: Robustness under IPM attack using a neural ranker. to evaluate incoming models. The same is the case for FLTrust and ZenoPS. However, RankGuard is cheaper per session: it only requires forward passes through the model, whereas FLTrust and ZenoPS replay sessions, involving PDGD gradient estimation and update steps. For example, at 100 sessions RankGuard requires roughly 130 ms, versus the 670 ms needed by F… view at source ↗
Figure 7
Figure 7. Figure 7: Attack detectability in relation to local session his [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Robustness under the Flip attack with attack share 𝛽 = 0.2. 0.2 0.4 WEB30K nDCG@10 perfect 0.2 0.4 navigational 0.2 0.4 informational 0.3 0.4 0.5 MQ2007 nDCG@10 0.2 0.4 0.2 0.4 0.5 0.6 0.7 Yahoo nDCG@10 0.5 0.6 0.7 0.5 0.6 0 5000 10000 15000 20000 25000 30000 Sessions 0.00 0.25 0.50 Istella nDCG@10 0 5000 10000 15000 20000 25000 30000 Sessions 0.00 0.25 0.50 0 5000 10000 15000 20000 25000 30000 Sessions 0.… view at source ↗
Figure 9
Figure 9. Figure 9: Robustness under the LIE attack with attack share 𝛽 = 0.2 [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Robustness under the IPM attack with attack share 𝛽 = 0.2 [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
read the original abstract

In Online Learning to Rank (OLTR), ranking models are trained directly from live user interactions, but existing systems rely on a trusted central server to collect and process these interactions. This leaves operators free to introduce biases that conflict with user interests. Decentralized learning offers an attractive alternative, allowing users to collaboratively train a shared ranking model by exchanging model updates directly with one another, without any central authority. In such settings, however, malicious nodes can send poisoned model updates that degrade the ranking quality of honest nodes. We introduce RankGuard, a decentralized OLTR framework in which users collaboratively train ranking models and exchange model updates directly with other nodes. RankGuard defends against poisoning attacks by carefully evaluating incoming models against the user's own private click history, corrected for position bias. An incoming model is only aggregated if it better explains the user's past interactions than the current local model, making it fundamentally hard for malicious nodes to craft updates that pass this test without also genuinely helping the user. We derive a theoretical convergence guarantee of RankGuard. To the best of our knowledge, this is the first formal convergence analysis of a decentralized OLTR algorithm. We evaluate RankGuard against four poisoning attacks, including a powerful adaptive attack, using four standard benchmarks and three click models. RankGuard outperforms all baselines in most settings while being up to 62x more efficient than its closest competitors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes RankGuard, a decentralized OLTR framework where nodes exchange model updates directly and aggregate an incoming model only if it yields higher likelihood on the user's private position-bias-corrected click history than the current local model. It claims this makes poisoning fundamentally difficult, derives a theoretical convergence guarantee (asserted to be the first formal analysis for decentralized OLTR), and reports empirical outperformance against four poisoning attacks (including adaptive) on four benchmarks and three click models, with up to 62x efficiency gains.

Significance. If the convergence guarantee holds under the stated assumptions and the filtering rule translates to genuine ranking-quality improvement, the result would be significant: it provides the first formal convergence analysis for decentralized OLTR and a practical defense against poisoning without a trusted server. The efficiency claims and multi-attack evaluation would further strengthen its contribution to robust decentralized collaborative ranking.

major comments (2)
  1. [Abstract and implied §3] Abstract and implied §3 (filtering rule): the claim that it is 'fundamentally hard for malicious nodes to craft updates that pass this test without also genuinely helping the user' rests on the position-bias-corrected history being an unbiased proxy for true preferences. The provided stress-test note correctly identifies that modest misspecification in the click model (examination probabilities, user-specific bias, or drift) could allow an adversary to maximize the corrected surrogate likelihood while degrading true NDCG; this assumption is load-bearing for both the robustness claim and the interpretation of the convergence guarantee.
  2. [Convergence analysis (abstract)] Convergence analysis (abstract): the guarantee is presented as derived rather than fitted, yet the abstract-only status and lack of an explicit statement on whether the bound applies to the corrected likelihood surrogate or to ranking quality under the real click process leaves open whether the result supports the central claim of genuine user benefit. A concrete test (e.g., relating the surrogate optimum to held-out NDCG) would be required to close this gap.
minor comments (1)
  1. The efficiency claim ('up to 62x more efficient') should specify the exact metric, baseline, and experimental conditions in the main text or a table for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive feedback on our manuscript. We address each major comment point by point below, indicating where revisions will be incorporated.

read point-by-point responses

  1. Referee: [Abstract and implied §3] Abstract and implied §3 (filtering rule): the claim that it is 'fundamentally hard for malicious nodes to craft updates that pass this test without also genuinely helping the user' rests on the position-bias-corrected history being an unbiased proxy for true preferences. The provided stress-test note correctly identifies that modest misspecification in the click model (examination probabilities, user-specific bias, or drift) could allow an adversary to maximize the corrected surrogate likelihood while degrading true NDCG; this assumption is load-bearing for both the robustness claim and the interpretation of the convergence guarantee.

    Authors: We agree that the filtering rule's effectiveness relies on the click model providing a reasonable proxy, and the manuscript already includes a stress-test note acknowledging potential misspecification effects. Our multi-benchmark, multi-click-model experiments demonstrate that RankGuard retains strong poisoning resistance and ranking performance in practice. We will revise the abstract and theory sections to more explicitly state the modeling assumptions and discuss their implications for robustness and convergence. revision: partial

  2. Referee: [Convergence analysis (abstract)] Convergence analysis (abstract): the guarantee is presented as derived rather than fitted, yet the abstract-only status and lack of an explicit statement on whether the bound applies to the corrected likelihood surrogate or to ranking quality under the real click process leaves open whether the result supports the central claim of genuine user benefit. A concrete test (e.g., relating the surrogate optimum to held-out NDCG) would be required to close this gap.

    Authors: The convergence guarantee is formally derived for the decentralized update process under the position-bias-corrected likelihood objective, following standard practice in OLTR theory. This surrogate is the natural objective for the algorithm. To address the gap, we will add an explicit clarification in the abstract and theory section, and include a new experiment relating surrogate likelihood to held-out NDCG in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: theoretical convergence claim is independent of fitted parameters

full rationale

The paper states that it derives a theoretical convergence guarantee for RankGuard as the first formal analysis of decentralized OLTR, with the aggregation decision defined directly from likelihood comparison on the user's position-bias-corrected private clicks. No equations or steps are shown that reduce the guarantee to a quantity defined by the same fitted parameters used in experiments, nor is any uniqueness theorem imported via self-citation. The derivation chain is presented as self-contained against external benchmarks rather than forced by construction or renaming of known results.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the position-bias correction and click models are treated as standard background rather than new inventions.

pith-pipeline@v0.9.1-grok · 5786 in / 1151 out tokens · 16585 ms · 2026-06-27T08:14:56.412152+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

70 extracted references · 10 canonical work pages

  1. [1]

    Alex Auvolat, Yérom-David Bromberg, Davide Frey, Djob Mvondo, and François Taïani. 2023. Basalt: A rock-solid byzantine-tolerant peer sampling for very large decentralized networks. InProceedings of the 24th International Middleware Conference. 111–123

    2023

  2. [2]

    Gilad Baruch, Moran Baruch, and Yoav Goldberg. 2019. A Little Is Enough: Circumventing Defenses For Distributed Learning. InAdvances in Neural Infor- mation Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, Hanna M. Wallach, Hugo Larochelle, Alina Beygelzimer, ...

    2019

  3. [3]

    Yacine Belal, Mohamed Maouche, Sonia Ben Mokhtar, and Anthony Simonet- Boulogne. 2025. Granite: a byzantine-resilient dynamic gossip learning frame- work.ArXiv preprintabs/2504.17471 (2025). https://arxiv.org/abs/2504.17471

    Pith/arXiv arXiv 2025

  4. [4]

    Yacine Belal, Mohamed Maouche, Sonia Ben Mokhtar, and Anthony Simonet- Boulogne. 2025. Inferring communities of interest in collaborative learning-based recommender systems. In2025 IEEE 45th International Conference on Distributed Computing Systems (ICDCS). IEEE, 505–515

    2025

  5. [5]

    Enrique Tomás Martínez Beltrán, Mario Quiles Pérez, Pedro Miguel Sánchez Sánchez, Sergio López Bernal, Gérôme Bovet, Manuel Gil Pérez, Grego- rio Martínez Pérez, and Alberto Huertas Celdrán. 2023. Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges.IEEE Communications Surveys & Tutorials25, 4 (2023), 2983–3013

    2023

  6. [6]

    Juan Benet. 2014. Ipfs-content addressed, versioned, p2p file system.arXiv preprint arXiv:1407.3561(2014)

    Pith/arXiv arXiv 2014

  7. [7]

    Sayan Biswas, Antoine Boutet, Davide Frey, Romaric Gaudel, Rachid Guerraoui, Maxime Jacovella, Anne-Marie Kermarrec, Dimitri Lerévérend, François Taïani, and Martijn de Vos. 2026. Your Neighbors Know: Leveraging Local Neighborhoods for Backdoor Detection in Decentralized Learning.arXiv preprint arXiv:2605.19969 (2026)

    Pith/arXiv arXiv 2026

  8. [8]

    Sayan Biswas, Mathieu Even, Anne-Marie Kermarrec, Laurent Massoulié, Rafael Pires, Rishi Sharma, and Martijn de Vos. 2025. Noiseless Privacy-Preserving Decentralized Learning.Proceedings on Privacy Enhancing Technologies(2025)

    2025

  9. [9]

    Sayan Biswas, Anne-Marie Kermarrec, Alexis Marouani, Rafael Pires, Rishi Sharma, and Martijn De Vos. 2025. Boosting asynchronous decentralized learning with model fragmentation. InProceedings of the ACM on Web Conference 2025. 685–696

    2025

  10. [10]

    Peva Blanchard, El Mahdi El Mhamdi, Rachid Guerraoui, and Julien Stainer. 2017. Machine Learning with Adversaries: Byzantine Tolerant Gradient Descent. InAd- vances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, December 4-9, 2017, Long Beach, CA, USA, Is- abelle Guyon, Ulrike von Luxburg, Sam...

    2017

  11. [11]

    Christopher J. C. Burges, Tal Shaked, Erin Renshaw, Ari Lazier, Matt Deeds, Nicole Hamilton, and Gregory N. Hullender. 2005. Learning to rank using gradient descent. InMachine Learning, Proceedings of the Twenty-Second International Conference (ICML 2005), Bonn, Germany, August 7-11, 2005 (ACM International Conference Proceeding Series, Vol. 119), Luc De ...

    work page doi:10.1145/1102351.1102363 2005

  12. [12]

    Xiaoyu Cao, Minghong Fang, Jia Liu, and Neil Zhenqiang Gong. 2021. FLTrust: Byzantine-robust Federated Learning via Trust Bootstrapping. In28th Annual Network and Distributed System Security Symposium, NDSS 2021, virtually, Febru- ary 21-25, 2021. The Internet Society. https://www.ndss-symposium.org/ndss- paper/fltrust-byzantine-robust-federated-learning-...

    2021

  13. [13]

    Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. 2007. Learning to rank: from pairwise approach to listwise approach. InMachine Learning, Proceedings of the Twenty-Fourth International Conference (ICML 2007), Corvallis, Oregon, USA, June 20-24, 2007 (ACM International Conference Proceeding Series, Vol. 227), Zoubin Ghahramani (Ed.). ACM, 129–13...

    work page doi:10.1145/1273496.1273513 2007

  14. [14]

    Olivier Chapelle and Yi Chang. 2011. Yahoo! learning to rank challenge overview. InProceedings of the learning to rank challenge. PMLR, 1–24

    2011

  15. [15]

    Olivier Chapelle and Ya Zhang. 2009. A dynamic bayesian network click model for web search ranking. InProceedings of the 18th International Conference on World Wide Web, WWW 2009, Madrid, Spain, April 20-24, 2009, Juan Quemada, Gonzalo León, Yoëlle S. Maarek, and Wolfgang Nejdl (Eds.). ACM, 1–10. doi:10. 1145/1526709.1526711

    arXiv 2009

  16. [16]

    European Commission. 2024. Commission opens formal proceedings against TikTok on election risks under the Digital Services Act. Press release IP/24/6487. Available from https://ec.europa.eu/commission/presscorner/detail/en/ip_24_ 6487

    2024

  17. [17]

    Domenico Dato, Sean MacAvaney, Franco Maria Nardini, Raffaele Perego, and Nicola Tonellotto. 2022. The Istella22 Dataset: Bridging Traditional and Neural Learning to Rank Evaluation. InSIGIR ’22: The 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Madrid, Spain, July 11 - 15, 2022, Enrique Amigó, Pablo Castell...

    arXiv 2022

  18. [18]

    Martijn de Vos, Sadegh Farhadkhani, Rachid Guerraoui, Anne-Marie Kermar- rec, Rafael Pires, and Rishi Sharma. 2023. Epidemic Learning: Boosting De- centralized Learning with Randomized Communication. InAdvances in Neural Information Processing Systems 36: Annual Conference on Neural Information Pro- cessing Systems 2023, NeurIPS 2023, New Orleans, LA, USA...

    2023

  19. [19]

    Trinh Viet Doan, Tat Dat Pham, Markus Oberprieler, and Vaibhav Bajpai. 2020. Measuring decentralized video streaming: A case study of dtube. In2020 IFIP Networking Conference (Networking). IEEE, 118–126

    2020

  20. [20]

    John R Douceur. 2002. The sybil attack. InInternational workshop on peer-to-peer systems. Springer, 251–260

    2002

  21. [21]

    Tim Draws, Nava Tintarev, Ujwal Gadiraju, Alessandro Bozzon, and Benjamin Timmermans. 2021. This Is Not What We Ordered: Exploring Why Biased Search Result Rankings Affect User Attitudes on Debated Topics. InSIGIR ’21: The 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event, Canada, July 11-15, 2021,...

    work page doi:10.1145/3404835.3462851 2021

  22. [22]

    El-Mahdi El-Mhamdi, Rachid Guerraoui, Arsany Guirguis, Lê Nguyên Hoang, and Sébastien Rouault. 2020. Genuinely distributed byzantine machine learning. InProceedings of the 39th Symposium on Principles of Distributed Computing. 355–364

    2020

  23. [23]

    Ahmed Roushdy Elkordy, Saurav Prakash, and Salman Avestimehr. 2022. Basil: A fast and Byzantine-resilient approach for decentralized training.IEEE Journal on Selected Areas in Communications40, 9 (2022), 2694–2716

    2022

  24. [24]

    Robert Epstein and Ronald E Robertson. 2015. The search engine manipulation effect (SEME) and its possible impact on the outcomes of elections.Proceedings of the national academy of sciences112, 33 (2015), E4512–E4521

    2015

  25. [25]

    Robert Epstein, Ronald E Robertson, David Lazer, and Christo Wilson. 2017. Suppressing the search engine manipulation effect (SEME).Proceedings of the ACM on Human-Computer Interaction1, CSCW (2017), 1–22

    2017

  26. [26]

    Mathieu Even, Anastasia Koloskova, and Laurent Massoulié. 2024. Asynchronous sgd on graphs: a unified framework for asynchronous decentralized and federated optimization. InInternational Conference on Artificial Intelligence and Statistics. PMLR, 64–72

    2024

  27. [27]

    Cheng Fang, Zhixiong Yang, and Waheed U Bajwa. 2022. BRIDGE: Byzantine- resilient decentralized gradient descent.IEEE Transactions on Signal and Infor- mation Processing over Networks8 (2022), 610–626

    2022

  28. [28]

    Minghong Fang, Xiaoyu Cao, Jinyuan Jia, and Neil Gong. 2020. Local model poisoning attacks to {Byzantine-Robust} federated learning. In29th USENIX security symposium (USENIX Security 20). 1605–1622

    2020

  29. [29]

    Minghong Fang, Zifan Zhang, Hairi, Prashant Khanduri, Jia Liu, Songtao Lu, Yuchen Liu, and Neil Gong. 2024. Byzantine-robust decentralized federated learning. InProceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security. 2874–2888

    2024

  30. [30]

    Chiara Farronato, Andrey Fradkin, and Alexander MacKay. 2023. Self- preferencing at Amazon: evidence from search rankings. InAEA Papers and Proceedings, Vol. 113. American Economic Association 2014 Broadway, Suite 305, Nashville, TN 37203, 239–243

    2023

  31. [31]

    Chao Feng, Alberto Huertas Celdrán, Janosch Baltensperger, Enrique Tomás Martínez Beltrán, Pedro Miguel Sánchez Sánchez, Gérôme Bovet, and Burkhard Stiller. 2024. Sentinel: An aggregation function to secure decentralized federated Preprint Gregoriadis et al. learning. InECAI 2024: 27th European Conference on Artificial Intelligence, 19– 24 October 2024, S...

    2024

  32. [32]

    Renaud Gaucher, Aymeric Dieuleveut, and Hadrien Hendrikx. 2024. Unified breakdown analysis for byzantine robust gossip.ArXiv preprintabs/2410.10418 (2024). https://arxiv.org/abs/2410.10418

    arXiv 2024

  33. [33]

    Amirata Ghorbani and James Zou. 2019. Data shapley: Equitable valuation of data for machine learning. InInternational conference on machine learning. PMLR, 2242–2251

    2019

  34. [34]

    Anindya Ghose, Panagiotis G Ipeirotis, and Beibei Li. 2014. Examining the impact of ranking on consumer behavior and search engine revenue.Management Science60, 7 (2014), 1632–1654

    2014

  35. [35]

    Marcel Gregoriadis, Rowdy M Chotkan, Petru Neague, and Johan Pouwelse. 2025. SwarmSearch: Decentralized Search Engine with Self-Funding Economy. In2025 IEEE 50th Conference on Local Computer Networks (LCN). IEEE, 1–10

    2025

  36. [36]

    Marcel Gregoriadis, Quinten Stokkink, and Johan Pouwelse. 2025. Decentralized adaptive ranking using transformers. InProceedings of the 5th Workshop on Machine Learning and Systems. 12–18

    2025

  37. [37]

    Rachid Guerraoui, Anne-Marie Kermarrec, Anastasiia Kucherenko, Rafael Pinot, and Martijn de Vos. 2024. Peerswap: A peer-sampler with randomness guarantees. In2024 43rd International Symposium on Reliable Distributed Systems (SRDS). IEEE, 271–281

    2024

  38. [38]

    Barbara Guidi, Andrea Michienzi, and Laura Ricci. 2022. Evaluating the decentral- isation of filecoin. InProceedings of the 3rd international workshop on distributed infrastructure for the common good. 13–18

    2022

  39. [39]

    Shangwei Guo, Tianwei Zhang, Han Yu, Xiaofei Xie, Lei Ma, Tao Xiang, and Yang Liu. 2021. Byzantine-resilient decentralized stochastic gradient descent.IEEE Transactions on Circuits and Systems for Video Technology32, 6 (2021), 4096–4106

    2021

  40. [40]

    Lie He, Sai Praneeth Karimireddy, and Martin Jaggi. 2022. Byzantine-robust decentralized learning via clippedgossip.ArXiv preprintabs/2202.01545 (2022). https://arxiv.org/abs/2202.01545

    arXiv 2022

  41. [41]

    Katja Hofmann, Shimon Whiteson, and Maarten de Rijke. 2013. Balancing ex- ploration and exploitation in listwise and pairwise online learning to rank for information retrieval.Information Retrieval16, 1 (2013), 63–90

    2013

  42. [42]

    Kalervo Järvelin and Jaana Kekäläinen. 2002. Cumulated gain-based evaluation of IR techniques.ACM Transactions on Information Systems (TOIS)20, 4 (2002), 422–446

    2002

  43. [43]

    Thorsten Joachims, Laura Granka, Bing Pan, Helene Hembrooke, and Geri Gay

  44. [44]

    InProceed- ings of the 28th Annual International ACM SIGIR Conference on Research and Devel- opment in Information Retrieval(Salvador, Brazil)(SIGIR ’05)

    Accurately interpreting clickthrough data as implicit feedback. InProceed- ings of the 28th Annual International ACM SIGIR Conference on Research and Devel- opment in Information Retrieval(Salvador, Brazil)(SIGIR ’05). Association for Com- puting Machinery, New York, NY, USA, 154–161. doi:10.1145/1076034.1076063

    work page doi:10.1145/1076034.1076063

  45. [45]

    Thorsten Joachims, Adith Swaminathan, and Tobias Schnabel. 2017. Unbiased Learning-to-Rank with Biased Feedback. InProceedings of the Tenth ACM Inter- national Conference on Web Search and Data Mining, WSDM 2017, Cambridge, United Kingdom, February 6-10, 2017, Maarten de Rijke, Milad Shokouhi, Andrew Tomkins, and Min Zhang (Eds.). ACM, 781–789. doi:10.114...

    work page doi:10.1145/3018661.3018699 2017

  46. [46]

    Eugene Kharitonov. 2019. Federated Online Learning to Rank with Evolution Strategies. InProceedings of the Twelfth ACM International Conference on Web Search and Data Mining, WSDM 2019, Melbourne, VIC, Australia, February 11-15, 2019, J. Shane Culpepper, Alistair Moffat, Paul N. Bennett, and Kristina Lerman (Eds.). ACM, 249–257. doi:10.1145/3289600.3290968

    work page doi:10.1145/3289600.3290968 2019

  47. [47]

    Anastasia Koloskova, Nicolas Loizou, Sadra Boreiri, Martin Jaggi, and Sebastian Stich. 2020. A unified theory of decentralized SGD with changing topology and local updates. InInternational conference on machine learning. PMLR, 5381–5393

    2020

  48. [48]

    Chang Li and Maarten de Rijke. 2019. Cascading Non-Stationary Bandits: Online Learning to Rank in the Non-Stationary Cascade Model. InProceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, IJCAI 2019, Macao, China, August 10-16, 2019, Sarit Kraus (Ed.). ijcai.org, 2859–2865. doi:10. 24963/ijcai.2019/396

    2019

  49. [49]

    Xiangru Lian, Ce Zhang, Huan Zhang, Cho-Jui Hsieh, Wei Zhang, and Ji Liu

  50. [50]

    Can Decentralized Algorithms Outperform Centralized Algorithms? A Case Study for Decentralized Parallel Stochastic Gradient Descent. InAdvances in Neu- ral Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, December 4-9, 2017, Long Beach, CA, USA, Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna M. Wa...

    2017

  51. [51]

    Tie-Yan Liu. 2010. Learning to rank for information retrieval. InProceeding of the 33rd International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2010, Geneva, Switzerland, July 19-23, 2010, Fabio Crestani, Stéphane Marchand-Maillet, Hsin-Hsi Chen, Efthimis N. Efthimiadis, and Jacques Savoy (Eds.). ACM, 904. doi:10.114...

    work page doi:10.1145/1835449.1835676 2010

  52. [52]

    Harrie Oosterhuis and Maarten de Rijke. 2018. Differentiable Unbiased On- line Learning to Rank. InProceedings of the 27th ACM International Conference on Information and Knowledge Management, CIKM 2018, Torino, Italy, October 22-26, 2018, Alfredo Cuzzocrea, James Allan, Norman W. Paton, Divesh Srivas- tava, Rakesh Agrawal, Andrei Z. Broder, Mohammed J. Z...

    work page doi:10.1145/3269206.3271686 2018

  53. [53]

    Róbert Ormándi, István Hegedűs, and Márk Jelasity. 2013. Gossip learning with linear models on fully distributed data.Concurrency and Computation: Practice and Experience25, 4 (2013), 556–571

    2013

  54. [54]

    Bing Pan, Helene Hembrooke, Thorsten Joachims, Lori Lorigo, Geri Gay, and Laura Granka. 2007. In Google we trust: Users’ decisions on rank, position, and relevance.Journal of computer-mediated communication12, 3 (2007), 801–823

    2007

  55. [55]

    Krishna Pillutla, Sham M Kakade, and Zaid Harchaoui. 2022. Robust aggregation for federated learning.IEEE Transactions on Signal Processing70 (2022), 1142– 1154

    2022

  56. [56]

    Frances A Pogacar, Amira Ghenai, Mark D Smucker, and Charles LA Clarke. 2017. The positive and negative influence of search results on people’s decisions about the efficacy of medical treatments. InProceedings of the ACM SIGIR International Conference on Theory of Information Retrieval. 209–216

    2017

  57. [57]

    Tao Qin and Tie-Yan Liu. 2013. Introducing LETOR 4.0 Datasets.CoRR abs/1306.2597 (2013). http://arxiv.org/abs/1306.2597

    Pith/arXiv arXiv 2013

  58. [58]

    Atul Singh, T-W Ngan, Peter Druschel, and Dan S Wallach. 2006. Eclipse attacks on overlay networks: Threats and defenses. InProceedings IEEE INFOCOM 2006. 25th IEEE international conference on computer communications

    2006

  59. [59]

    Ousmane Touat, Jezekael Brunon, Yacine Belal, Julien Nicolas, César Sabater, Mohamed Maouche, and Sonia Ben Mokhtar. 2025. Exposing the Vulnerability of Decentralized Learning to Membership Inference Attacks Through the Lens of Graph Mixing. InProceedings of the 26th International Middleware Conference. 180–194

    2025

  60. [60]

    Raluca M Ursu. 2018. The power of rankings: Quantifying the effect of rankings on online consumer search and purchase decisions.Marketing Science37, 4 (2018), 530–552

    2018

  61. [61]

    Shuyi Wang, Bing Liu, Shengyao Zhuang, and Guido Zuccon. 2021. Effective and privacy-preserving federated online learning to rank. InProceedings of the 2021 ACM SIGIR international conference on theory of information retrieval. 3–12

    2021

  62. [62]

    Shuyi Wang and Guido Zuccon. 2023. An analysis of untargeted poisoning attack and defense methods for federated online learning to rank systems. InProceedings of the 2023 ACM SIGIR International Conference on Theory of Information Retrieval. 215–224

    2023

  63. [63]

    Stefanie Warnat-Herresthal, Hartmut Schultze, Krishnaprasad Lingadahalli Shastry, Sathyanarayanan Manamohan, Saikat Mukherjee, Vishesh Garg, Ravi Sarveswara, Kristian Händler, Peter Pickkers, N Ahmad Aziz, et al. 2021. Swarm learning for decentralized and confidential clinical machine learning.Nature594, 7862 (2021), 265–270

    2021

  64. [64]

    Cong Xie, Oluwasanmi Koyejo, and Indranil Gupta. 2019. Fall of Empires: Breaking Byzantine-tolerant SGD by Inner Product Manipulation. InProceed- ings of the Thirty-Fifth Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel A viv, Israel, July 22-25, 2019 (Proceedings of Machine Learning Re- search, Vol. 115), Amir Globerson and Ricardo Sil...

    2019

  65. [65]

    Cong Xie, Oluwasanmi Koyejo, and Indranil Gupta. 2022. Zenops: A distributed learning system integrating communication efficiency and security.Algorithms 15, 7 (2022), 233

    2022

  66. [66]

    Cong Xie, Sanmi Koyejo, and Indranil Gupta. 2020. Zeno++: Robust Fully Asyn- chronous SGD. InProceedings of the 37th International Conference on Machine Learning, ICML 2020, 13-18 July 2020, Virtual Event (Proceedings of Machine Learn- ing Research, Vol. 119). PMLR, 10495–10503. http://proceedings.mlr.press/v119/ xie20c.html

    2020

  67. [67]

    Caiyi Yang and Javad Ghaderi. 2024. Byzantine-Robust Decentralized Learning via Remove-then-Clip Aggregation. InThirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Conference on Innovative Applications of Artificial Intelligence, IAAI 2024, Fourteenth Symposium on Educational Advances in Artificial Intelligence, EAAI 2014, Fe...

    work page doi:10.1609/aaai.v38i19.30173 2024

  68. [68]

    Zhixiong Yang and Waheed U Bajwa. 2019. ByRDiE: Byzantine-resilient dis- tributed coordinate descent for decentralized learning.IEEE Transactions on Signal and Information Processing over Networks5, 4 (2019), 611–627

    2019

  69. [69]

    Bartlett

    Dong Yin, Yudong Chen, Kannan Ramchandran, and Peter L. Bartlett. 2018. Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates. In Proceedings of the 35th International Conference on Machine Learning, ICML 2018, Stockholmsmässan, Stockholm, Sweden, July 10-15, 2018 (Proceedings of Machine Learning Research, Vol. 80), Jennifer G. Dy and A...

    2018

  70. [70]

    Yisong Yue and Thorsten Joachims. 2009. Interactively optimizing informa- tion retrieval systems as a dueling bandits problem. InProceedings of the 26th Annual International Conference on Machine Learning, ICML 2009, Montreal, Que- bec, Canada, June 14-18, 2009 (ACM International Conference Proceeding Series, Vol. 382), Andrea Pohoreckyj Danyluk, Léon Bot...

    work page doi:10.1145/1553374.1553527 2009