Pith. sign in

REVIEW 4 major objections 7 minor 2 cited by

Federated Deconfounding and Debiasing Learning for Out-of-Distribution Generalization

T0 review · 4 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read FedDDL claims that removing background and source confounders from federated inference raises out-of-distribution accuracy by 4.5% on average.

desk verdict A plausible federated OOD recipe whose central causal deconfounding claim does not survive contact with its own equations; deserves peer review with major revision, not desk rejection. read the letter →

arxiv 2505.04979 v2 pith:MERR7GAW submitted 2025-05-08 cs.CV

Zhuang Qi , Sijin Zhou , Lei Meng , Han Hu , Han Yu , Xiangxu Meng This is my paper · ORCID
classification cs.CV
keywords federatedlearningout-of-distributiongeneralizationdeconfoundingbackdooradjustmentcounterfactualsamplescausalprototypesbackgroundbiasobjectdetection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that federated models fail on out-of-distribution data because they learn two spurious inference paths: a background factor $B$ that correlates with labels within each client, and a source factor $S$ that differs between clients. It proposes FedDDL, which builds a structural causal graph over these factors and performs backdoor adjustment to block both paths. An intra-client module splits each image into object and background with a text-prompted detector, then fuses the object with random backgrounds so the model cannot lean on scenery. An inter-client module constructs prototypes from object images alone and aligns local features to them, shrinking cross-client output gaps. If the causal repair works as described, it explains the reported 4.5% average Top-1 accuracy improvement over nine baselines on NICO-Animal and NICO-Vehicle.

What carries the argument

The load-bearing machinery is the structural causal graph over Source $S$, Background $B$, Object $O$, Image $X$, and Label $Y$, together with the backdoor adjustment formula $P(Y|\mathrm{do}(X)) = \sum_B P(Y|X,B)P(B)$. The mechanism runs through a pre-trained open-set object detector (Grounding DINO) prompted with the class name to produce a bounding box; Hadamard masks split each image into object image $I_O$ and background image $I_B$; counterfactual samples $I_C$ are created by fusing $I_O$ with randomly grouped backgrounds from the same client; and causal prototypes are class-wise mean embeddings of $I_O$ computed by the global model, used in a contrastive loss that pulls local features toward same-class prototypes and away from other-class prototypes. All of this is designed to make the background contribute equally across labels and to align heterogeneous client representations so that aggregation does not carry spurious correlations.

What would settle it

Train FedDDL on a version of NICO where the object detector is deliberately weakened (e.g., small, occluded, or camouflaged objects with lower detection confidence): if the 4.5% average gain collapses below FedAvg's accuracy while all other components remain unchanged, the claimed deconfounding is not what drives the improvement.

Watch

Extended reading notes

Core claim

The central claim is that federated out-of-distribution generalization fails through two identifiable confounding paths, $X \leftarrow B \to Y$ and $X \leftarrow S \to Y$, and that both can be severed without exchanging raw samples. FedDDL replaces the confounded inference $P(Y|X)$ with the backdoor-adjusted $P(Y|\mathrm{do}(X))$ by generating counterfactual images whose backgrounds are decoupled from labels and by regularizing local representations against causal prototypes computed from object regions alone. The result is that models learn to predict from the object feature $O$, so unseen backgrounds no longer hijack the decision. The paper presents this as both a causal analysis of the inference paths and a method, with the 4.5% average Top-1 gain as evidence that the deconfounding, not mere data augmentation, drives the improvement.

Load-bearing premise

The method stands on the assumption that the text-prompted object detector reliably separates the truly causal object from the background in every client's images, so the counterfactual backgrounds and causal prototypes carry no hidden spurious information.

Editorial extensions

If this is right

  • FedDDL reports 4.5% higher Top-1 accuracy on average than nine baselines across NICO-Animal and NICO-Vehicle, improving both global and averaged local models.
  • The DEC module works as a plug-in: adding it to MOON (MOON+DEC) improves that baseline substantially, suggesting the counterfactual intervention transfers across federated learning methods.
  • The DEB module reduces performance disparity among local clients by aligning their representations to shared causal prototypes, as shown by lower local accuracy variance and t-SNE alignment.
  • Ablation on the L7 splits shows the gains of DEC and DEB are additive: full FedDDL exceeds either module alone, and benefit grows with the number of counterfactual background groups $\eta$.
  • With a fixed ResNet-18 backbone, FedDDL remains above FedAvg across the tested hyperparameter ranges $\lambda \in \{0.1, 1.0, 2.0\}$, $\tau \in \{0.07, 0.5\}$, and $\eta \in \{1, 3, 5\}$.

Reading between the lines

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

  • If the causal story is right, FedDDL should also improve robustness to background shifts not present in NICO's seven training backgrounds, for example by evaluating on held-out scenes from other datasets, which the paper does not test.
  • Because counterfactual backgrounds are mixed only within each client's own images, the method preserves the privacy boundary of federated learning more strictly than style-transfer augmentation that exchanges style statistics across clients; this is an implied benefit, not a claim the paper tests.
  • A natural extension, which the paper names as future work, is replacing the fixed text-prompted detector with learned or adaptive object localization when attributes cannot be directly disentangled, such as in medical images.
  • The method's object-background split is binary; extending it to multi-part object structure (e.g., animal head versus body versus habitat) would test whether the deconfounding gain comes from removing background only or from isolating the most discriminative sub-region.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. This paper proposes FedDDL, a federated learning method for out-of-distribution generalization. The authors build a structural causal model in which image background B and data source S act as confounders of the image X and label Y, and they claim that an intra-client deconfounding module (DEC) implements backdoor adjustment by generating counterfactual images with recombined backgrounds, while an inter-client debiasing module (DEB) aligns client representations through class-wise causal prototypes computed from object-only images under a contrastive loss. The method is evaluated on NICO-Animal and NICO-Vehicle in terms of global and local Top-1 accuracy against nine baselines, with ablations, hyperparameter sensitivity studies, and qualitative attention analyses; the reported average improvement over the best baseline is about 4.5 percentage points.

Significance. The empirical component of the paper is solid: the evaluation uses standard benchmarks with external test sets and held-out backgrounds, nine baselines, three-run statistics, a module-level ablation (Table 3), hyperparameter sensitivity (Fig. 6), a plug-and-play demonstration (MOON+DEC in Table 2), and honest reporting of local-model variance. The global-model gains are large relative to the reported standard deviations, and the claimed average improvement of roughly 4.5% over the strongest baseline is consistent with Table 2. If the deconfounding interpretation were correct, this would be a principled new perspective for federated OOD generalization. However, as specified, the DEC module is an intra-class background-mixing data augmentation and the DEB module is a prototype-based contrastive regularizer; the paper does not establish that either procedure implements the backdoor adjustment in Eqs. (2) and (9). The causal narrative is therefore not supported by the described algorithm, and the durable contribution is a well-evaluated empirical method whose causal interpretation needs either a corrected specification or a softened claim.

major comments (4)
  1. [§4.2, Eqs. (5)–(7)] The counterfactual generation procedure as written does not randomize backgrounds across labels, so it does not produce the interventional distribution P(Y|do(X)) claimed in Eq. (2). Eq. (5) groups backgrounds I_i^B 'belonging to class i' into class-specific groups g_j^i, and Eq. (6) fuses those same-class background groups with the object I_O; the classification loss in Eq. (7) then assigns the original true class label to each counterfactual sample. Under the most direct reading, every counterfactual is a class-i image with a class-i background, so the background-label co-occurrence in D_C,k mirrors the training data and the spurious B→Y path is preserved. The abstract's claim that the counterfactuals 'establish a connection between the background and any label' would require pairing background groups sampled from all classes with objects of all classes, but that operation is not described in the method text or in Algorithm 1. Please specify the cross-class pairing rule (including the label rule for mixed pairs), or alternatively revise the causal-intervention claim.
  2. [§4.2, Eqs. (2) and (6)] Even under the intended cross-class pairing, the paper gives no argument that the generated counterfactual distribution matches Σ_B P(Y|X,B)P(B). The Split(·) operator in Eq. (5) creates η class-indexed groups, and Eq. (6) averages the images within each group, so the counterfactual backgrounds are class-conditional averages rather than samples from the background marginal P(B). In addition, the two terms of Eq. (7) are added with equal weight, which does not correspond to any stated weighting by P(B). Without an explicit account of how the counterfactual sampling distribution relates to the intervention in Eq. (2), the equality connecting the backdoor formula to the DEC loss is not established.
  3. [§4.3, Eqs. (9)–(12)] The DEB module is described as performing the interventional computation in Eq. (9), but the module actually computes class-wise prototypes from object crops (Eq. 10), averages them across clients (Eq. 11), and applies a supervised contrastive loss (Eq. 12). No term in L_CR corresponds to the summation over P(B)P(S_i) in Eq. (9); P(S_i) is never defined for the fixed, finite set of FL clients, and the text does not explain why aligning local features to object-based prototypes instantiates the backdoor adjustment. The module's actual effect—cross-client representation alignment—is a legitimate technical goal and should be claimed directly rather than through Eq. (9).
  4. [§5.1, Eqs. (3)–(4)] The entire method rests on Grounding DINO's bounding boxes, with the BOX THRESHOLD and TEXT THRESHOLD fixed at 0.3, but the paper reports no measure of detection quality (e.g., per-class detection rate or object/background separation quality) and no sensitivity analysis over these thresholds. If the detector misses small or occluded objects, then I_B still contains the causal object and I_O can be near-empty, which would corrupt both the counterfactual samples of Eq. (6) and the causal prototypes of Eq. (10). A quantitative check of the detector's reliability is needed to support the claim that the object/background decomposition is sound.
minor comments (7)
  1. [Eq. (7)] The second term of Eq. (7) is garbled ('+ −Σ_{C}^{n=1} y_C(n) log(...)'); it should be written over a counterfactual-sample index with a consistent summation bound, and the relative weight of the two loss terms should be stated explicitly.
  2. [Algorithm 1, lines 16–19] Line 18 aggregates θ^t as the average of the previous-round parameters θ^{t−1}_k rather than the updated local models θ^t_k, and the pseudocode's uniform 1/|K| weighting is inconsistent with the dataset-size weights α_k defined in Section 3; please clarify which aggregation rule was used in the experiments.
  3. [§4.1 and Figure 3 caption] The Figure 3 caption states that the modules sever 'B ↛ X' and 'S ↛ X', while Section 4.1 says the modules sever 'X ↚ B' and 'X ↚ S' and the path B→Y; the target of the claimed intervention should be stated consistently throughout.
  4. [Table 2] The baseline name appears as 'FedDeccor' in Table 2 but 'FedDecorr' in Section 5.2; please make the spelling consistent.
  5. [§5.3] The phrase 'decbiasing learning module' appears twice in the ablation discussion and should read 'debiasing learning module'.
  6. [§5.1, Implementation Details] The text contains an unresolved placeholder citation, 'Following prior studies [?; Liu et al., 2021]'; the missing reference should be supplied.
  7. [References] Several entries use the nonstandard form '[Qi and et al., 2024]' and '[Zhang and et al., 2024]' with the author list replaced by 'and et al.', and the names in a few entries (e.g., '´Slazyk') contain stray accent artifacts; the bibliography should be brought into the venue's format.

Circularity Check

0 steps flagged · score 2.0 of 10

No meaningful circularity: FedDDL's causal graph and counterfactual construction are explicit inputs, and its accuracy claims are checked against external NICO test data and independent baselines.

full rationale

The central claim is not derived by fitting a parameter to the test set and then relabeling the fit as a prediction. The method's causal graph (Figure 2), the backdoor-adjustment equations (Eqs. 1-2, 8-9), and the two modules (DEC and DEB) are all constructed in the paper rather than imported from a self-citation chain. The reported 4.5% average Top-1 improvement is evaluated on the external NICO-Animal and NICO-Vehicle benchmarks against nine baselines, so the evaluation is self-contained outside the fitted values. The self-citations in the introduction and related work (e.g., Qi et al. 2024, Qi et al. 2025b, Meng et al. 2024) are contextual references and are not load-bearing: no uniqueness theorem, intervening variable, or core derivation is justified solely by those citations. One legitimate concern, raised by the skeptic, is that the counterfactual construction as written in Eqs. 5-6 may not randomize backgrounds across classes, in which case the claimed B-perp-Y intervention would not be achieved. That is a correctness or implementation ambiguity, not a circular reduction: the paper's equations do not equate the counterfactual distribution to the test result, nor does the reported accuracy reduce to an input by construction. There is also a dangling citation placeholder in Section 5.1 ('Following prior studies [?; Liu et al., 2021]'), which is an editorial completeness issue rather than a circularity issue. Overall, no circular step is established, so the score is low.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The method relies on a hand-specified causal graph and a pretrained detector; the free hyperparameters are tuned on validation (values not reported). No new physical entities are introduced.

free parameters (4)
  • lambda (contrastive loss weight) = not stated; selected from {0.1, 1.0, 2.0}
    Tuned on validation; final value not reported in paper.
  • tau (temperature) = not stated; selected from {0.07, 0.5}
    Tuned on validation; paper notes 0.07 improves separation.
  • eta (number of background groups) = not stated; selected from {1, 3, 5}
    Controls diversity of counterfactual backgrounds; tuned on validation.
  • DINO box/text threshold = 0.3
    Chosen by hand; no sensitivity analysis reported.
assumptions (3)
  • domain assumption The causal graph X <- B -> Y and X <- S -> Y is complete for federated image inference (Section 4.1).
    The paper assumes background B and source S are the only confounders; other attributes like pose, color, or illumination are ignored.
  • standard math Backdoor adjustment is valid given the assumed causal graph (Eq. 2, Eq. 9).
    The formulas follow from the graph, but the paper does not prove the graph structure from data.
  • domain assumption Grounding DINO provides correct object masks across all clients (Eq. 3-4).
    The entire DEC module depends on reliable object detection; failure cases are not analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Federated Deconfounding and Debiasing Learning for Out-of-Distribution Generalization." pith.science (2026). https://pith.science/paper/MERR7GAW

@misc{pith2026250504979,
  author       = {Pith},
  title        = {Pith review of: Federated Deconfounding and Debiasing Learning for Out-of-Distribution Generalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MERR7GAW}},
  note         = {Machine review of arXiv:2505.04979}
}
read the original abstract

Attribute bias in federated learning (FL) typically leads local models to optimize inconsistently due to the learning of non-causal associations, resulting degraded performance. Existing methods either use data augmentation for increasing sample diversity or knowledge distillation for learning invariant representations to address this problem. However, they lack a comprehensive analysis of the inference paths, and the interference from confounding factors limits their performance. To address these limitations, we propose the \underline{Fed}erated \underline{D}econfounding and \underline{D}ebiasing \underline{L}earning (FedDDL) method. It constructs a structured causal graph to analyze the model inference process, and performs backdoor adjustment to eliminate confounding paths. Specifically, we design an intra-client deconfounding learning module for computer vision tasks to decouple background and objects, generating counterfactual samples that establish a connection between the background and any label, which stops the model from using the background to infer the label. Moreover, we design an inter-client debiasing learning module to construct causal prototypes to reduce the proportion of the background in prototype components. Notably, it bridges the gap between heterogeneous representations via causal prototypical regularization. Extensive experiments on 2 benchmarking datasets demonstrate that \methodname{} significantly enhances the model capability to focus on main objects in unseen data, leading to 4.5\% higher Top-1 Accuracy on average over 9 state-of-the-art existing methods.

Figures

Figures reproduced from arXiv: 2505.04979 by the authors.

Figure 1
Figure 1. FedDDL reveals two factors affecting model inference: specific backgrounds and the output gap between data sources. (a) It constructs a structured causal graph to support the claim. (b) Dogs Samples illustrates the specific background within the client. (c) It represents the feature distribution differences between data sources. train models locally and iteratively exchange parameters with the server, enabling globa… view at source ↗
Figure 2
Figure 2. A causal view in federated out-of-distribution generaliza [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The FedDDL framework. It contains two main modules: 1) the intra-client deconfounding learning module, and 2) the inter-client debiasing learning module. The former generates counterfactual samples to break the spurious association between the background and specific labels B ↛ X. The latter leverages causal prototypes to promote consistency in learning heterogeneous representations S ↛ X. and background (B → X) fac… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Example counterfactual samples with causal features. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Comparison between traditional and causal prototypes. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: The impact of hyperparameters on performance. The [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 9
Figure 9. Figure 9: Visualization of representation distributions from differ [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 8
Figure 8. Figure 8: Visualization of visual attention and comparison of pre [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Every Client Is an Environment: Federated De-confounding for Spatio-Temporal Forecasting

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Federated aggregation of aligned environmental prototype codebooks acts as implicit back-door adjustment, with de-confounding error bounded linearly by average client confounding strength.

  2. Out-of-Distribution Federated Distillation with Domain-Aware Proxy

    cs.LG 2026-08 reject novelty 4.0 of 10

    DPS-FD, a domain-aware proxy selection method with a vocabulary-constrained LLM proxy generator, is claimed to improve out-of-distribution federated distillation, but its OOD evaluation is compromised by proxy data dr...

Reference graph

Works this paper leans on

53 extracted references · 48 canonical work pages · cited by 2 Pith papers

  1. [3]

    Mit- igating data heterogeneity in federated learning with data augmentation

    [de Luca and et al., 2022] Artur Back de Luca and et al. Mit- igating data heterogeneity in federated learning with data augmentation. arXiv preprint arXiv:2206.09979,

    arXiv 2022

  2. [6]

    Deep residual learning for image recog- nition

    [He et al., 2016] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recog- nition. In CVPR, pages 770–778,

    2016

  3. [8]

    Fedmut: Generalized federated learning via stochastic mu- tation

    [Hu et al., 2024] Ming Hu, Zeke Xia, Dengke Yan, and et al. Fedmut: Generalized federated learning via stochastic mu- tation. In AAAI, volume 38, pages 12528–12537,

    work page 2024

  4. [9]

    Re- thinking federated learning with domain shift: A prototype view

    [Huang et al., 2023] Wenke Huang, Mang Ye, and et al. Re- thinking federated learning with domain shift: A prototype view. In CVPR, pages 16312–16322. IEEE,

    work page 2023

  5. [10]

    A cross-client coordinator in fed- erated learning framework for conquering heterogeneity

    [Huang et al., 2025] Sheng Huang, Lele Fu, Yuecheng Li, Chuan Chen, and et al. A cross-client coordinator in fed- erated learning framework for conquering heterogeneity. IEEE Transactions on Neural Networks and Learning Sys- tems, 36(5):8828–8842,

    work page 2025

  6. [11]

    Feder- ated optimization in heterogeneous networks.Proceedings of Machine learning and systems, 2:429–450,

    [Li et al., 2020] Tian Li, Anit Kumar Sahu, and et al. Feder- ated optimization in heterogeneous networks.Proceedings of Machine learning and systems, 2:429–450,

    work page 2020

  7. [14]

    Multi-source domain adaptation for visual sentiment classification

    [Lin et al., 2020] Chuang Lin, Sicheng Zhao, Lei Meng, and Tat-Seng Chua. Multi-source domain adaptation for visual sentiment classification. In AAAI, volume 34, pages 2661– 2668,

    work page 2020

  8. [15]

    Grounding dino: Marrying dino with grounded pre-training for open-set ob- ject detection

    [Liu and et al., 2025] Shilong Liu and et al. Grounding dino: Marrying dino with grounded pre-training for open-set ob- ject detection. In ECCV, pages 38–55. Springer,

    work page 2025

Show all 53 references
  1. [16]

    Feddg: Federated domain generalization on medi- cal image segmentation via episodic learning in continu- ous frequency space

    [Liu et al., 2021] Quande Liu, Cheng Chen, Jing Qin, and et al. Feddg: Federated domain generalization on medi- cal image segmentation via episodic learning in continu- ous frequency space. In CVPR, pages 1013–1023,

    2021

  2. [17]

    Disentangled federated learning for tack- ling attributes skew via invariant aggregation and diversity transferring

    [Luo et al., 2022] Zhengquan Luo, Yunlong Wang, Zilei Wang, and et al. Disentangled federated learning for tack- ling attributes skew via invariant aggregation and diversity transferring. arXiv preprint arXiv:2206.06818,

    2022 arXiv

  3. [18]

    Communication-efficient learn- ing of deep networks from decentralized data

    [McMahan et al., 2017] Brendan McMahan, Eider Moore, Daniel Ramage, and et al. Communication-efficient learn- ing of deep networks from decentralized data. InAISTATS, pages 1273–1282. PMLR,

    2017

  4. [21]

    Stable diffusion-based data augmentation for federated learning with non-iid data

    [Morafah and et al., 2024] Mahdi Morafah and et al. Stable diffusion-based data augmentation for federated learning with non-iid data. arXiv preprint arXiv:2405.07925,

    2024 arXiv

  5. [22]

    Fedproc: Prototypical contrastive federated learning on non-iid data

    [Mu et al., 2023] Xutong Mu, Yulong Shen, Ke Cheng, Xueli Geng, and et al. Fedproc: Prototypical contrastive federated learning on non-iid data. Future Generation Computer Systems, 143:93–104,

    2023

  6. [23]

    Fisc: Federated domain generaliza- tion via interpolative style transfer and contrastive learn- ing

    [Nguyen et al., 2024] Dung Thuy Nguyen, Taylor T John- son, and Kevin Leach. Fisc: Federated domain generaliza- tion via interpolative style transfer and contrastive learn- ing. arXiv preprint arXiv:2410.22622,

    2024 arXiv

  7. [24]

    Stablefdg: style and attention based learning for federated domain generalization

    [Park et al., 2024] Jungwuk Park, Dong-Jun Han, and et al. Stablefdg: style and attention based learning for federated domain generalization. NeurIPS, 36,

    2024

  8. [25]

    Attentive model- ing and distillation for out-of-distribution generalization of federated learning

    [Qi and et al., 2024] Zhuang Qi and et al. Attentive model- ing and distillation for out-of-distribution generalization of federated learning. In ICME, pages 1–6. IEEE,

    2024

  9. [26]

    Clustering-based curriculum construction for sample-balanced federated learning

    [Qi et al., 2022] Zhuang Qi, Yuqing Wang, Zitan Chen, Ran Wang, Xiangxu Meng, and Lei Meng. Clustering-based curriculum construction for sample-balanced federated learning. In CICAI, pages 155–166. Springer,

    2022

  10. [27]

    Cross-silo prototypical calibration for federated learning with non-iid data

    [Qi et al., 2023] Zhuang Qi, Lei Meng, Zitan Chen, and et al. Cross-silo prototypical calibration for federated learning with non-iid data. In MM, pages 3099–3107,

    2023

  11. [28]

    Cross-training with multi-view knowl- edge fusion for heterogenous federated learning

    [Qi et al., 2024] Zhuang Qi, Lei Meng, Weihao He, Ruohan Zhang, and et al. Cross-training with multi-view knowl- edge fusion for heterogenous federated learning. arXiv preprint arXiv:2405.20046, pages 1–12,

    2024

  12. [29]

    Advances and open chal- lenges in federated foundation models

    [Ren et al., 2025] Chao Ren, Han Yu, Hongyi Peng, Xiaoli Tang, Bo Zhao, Liping Yi, et al. Advances and open chal- lenges in federated foundation models. IEEE Communica- tions Surveys and Tutorials,

    2025

  13. [30]

    Grad-cam: Visual explanations from deep networks via gradient-based local- ization

    [Selvaraju et al., 2017] Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, and et al. Grad-cam: Visual explanations from deep networks via gradient-based local- ization. In ICCV, pages 618–626,

    2017

  14. [31]

    Learning across domains and devices: Style-driven source-free domain adaptation in clustered federated learning

    [Shenaj et al., 2023] Donald Shenaj, Eros Fan `ı, Marco Toldo, and et al. Learning across domains and devices: Style-driven source-free domain adaptation in clustered federated learning. In WACV, pages 444–454,

    2023

  15. [32]

    Understanding and mitigating di- mensional collapse in federated learning

    [Shi et al., 2024] Yujun Shi, Jian Liang, Wenqing Zhang, Chuhui Xue, and et al. Understanding and mitigating di- mensional collapse in federated learning. IEEE Trans- actions on Pattern Analysis and Machine Intelligence , 46(5):2936–2949,

    2024

  16. [33]

    Cxr-fl: deep learning-based chest x-ray image analysis using federated learning

    [´Slazyk et al., 2022] Filip ´Slazyk, Przemysław Jabłecki, Aneta Lisowska, Maciej Malawski, and et al. Cxr-fl: deep learning-based chest x-ray image analysis using federated learning. In ICCS, pages 433–440. Springer,

    2022

  17. [34]

    Feature distribution matching for federated domain gener- alization

    [Sun et al., 2023] Yuwei Sun, Ng Chong, and Hideya Ochiai. Feature distribution matching for federated domain gener- alization. In ACML, pages 942–957. PMLR,

    2023

  18. [35]

    Visualizing data using t-sne

    [Van der Maaten and Hinton, 2008] Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. J MACH LEARN RES, 9(11),

    2008

  19. [37]

    Causal attention for unbiased visual recognition

    [Wang et al., 2021] Tan Wang, Chang Zhou, Qianru Sun, and Hanwang Zhang. Causal attention for unbiased visual recognition. In CVPR, pages 3091–3100,

    2021

  20. [38]

    Meta-causal feature learning for out-of-distribution gener- alization

    [Wang et al., 2022] Yuqing Wang, Xiangxian Li, and et al. Meta-causal feature learning for out-of-distribution gener- alization. In ECCV, pages 530–545. Springer,

    2022

  21. [39]

    Dafkd: Domain-aware federated knowledge distillation

    [Wang et al., 2023] Haozhao Wang, Yichen Li, Wenchao Xu, and et al. Dafkd: Domain-aware federated knowledge distillation. In CVPR, pages 20412–20421,

    2023

  22. [40]

    Fedcda: Federated learning with cross-rounds divergence-aware aggregation

    [Wang et al., 2024] Haozhao Wang, Haoran Xu, Yichen Li, and et al. Fedcda: Federated learning with cross-rounds divergence-aware aggregation. In ICLR,

    2024

  23. [41]

    Multi- source collaborative gradient discrepancy minimization for federated domain generalization

    [Wei and Han, 2024] Yikang Wei and Yahong Han. Multi- source collaborative gradient discrepancy minimization for federated domain generalization. In AAAI, volume 38, pages 15805–15813,

    2024

  24. [42]

    Federated adversarial domain halluci- nation for privacy-preserving domain generalization.IEEE Transactions on Multimedia, 26:1–14,

    [Xu et al., 2023] Qinwei Xu, Ruipeng Zhang, Ya Zhang, Yi- Yan Wu, and et al. Federated adversarial domain halluci- nation for privacy-preserving domain generalization.IEEE Transactions on Multimedia, 26:1–14,

    2023

  25. [43]

    Fed- erated learning: Privacy and incentive,

    [Yang et al., 2020] Qiang Yang, Lixin Fan, and Han Yu. Fed- erated learning: Privacy and incentive,

    2020

  26. [44]

    Fedgh: Heterogeneous federated learning with generalized global header

    [Yi et al., 2023] Liping Yi, Gang Wang, and et al. Fedgh: Heterogeneous federated learning with generalized global header. In MM, pages 8686–8696. ACM,

    2023

  27. [45]

    Fed- erated model heterogeneous matryoshka representation learning

    [Yi et al., 2024] Liping Yi, Han Yu, Chao Ren, et al. Fed- erated model heterogeneous matryoshka representation learning. In NeurIPS,

    2024

  28. [46]

    Dynamic witness se- lection for trustworthy distributed cooperative sensing in cognitive radio networks

    [Yu and et al., 2011] Han Yu and et al. Dynamic witness se- lection for trustworthy distributed cooperative sensing in cognitive radio networks. In ICCT, pages 1–6,

    2011

  29. [48]

    Enabling collaborative test-time adaptation in dynamic environment via federated learning

    [Zhang and et al., 2024] Jiayuan Zhang and et al. Enabling collaborative test-time adaptation in dynamic environment via federated learning. In KDD, pages 4191–4202,

    2024

  30. [49]

    Federated analytics with data augmentation in domain generalization towards future networks

    [Zhang et al., 2024] Xunzheng Zhang, Juan Marcelo Parra- Ullauri, Shadi Moazzeni, and et al. Federated analytics with data augmentation in domain generalization towards future networks. IEEE Transactions on Machine Learning in Communications and Networking,

    2024

  31. [50]

    Federated out-of-distribution generaliza- tion: A causal augmentation view

    [Zhang et al., 2025b] Runhui Zhang, Sijin Zhou, and Zhuang Qi. Federated out-of-distribution generaliza- tion: A causal augmentation view. arXiv preprint arXiv:2504.19882,

    arXiv

  32. [51]

    Fed- erated learning based on diffusion model to cope with non- iid data

    [Zhao et al., 2023] Zhuang Zhao, Feng Yang, and et al. Fed- erated learning based on diffusion model to cope with non- iid data. In PRCV, pages 220–231. Springer,

    2023

  33. [52]

    Fedfa: Federated learning with feature anchors to align features and classifiers for heterogeneous data

    [Zhou et al., 2023] Tailin Zhou, Jun Zhang, and Danny HK Tsang. Fedfa: Federated learning with feature anchors to align features and classifiers for heterogeneous data. IEEE Transactions on Mobile Computing,

    2023

  34. [53]

    Dualfed: enjoying both generalization and per- sonalization in federated learning via hierachical represen- tations

    [Zhu et al., 2024] Guogang Zhu, Xuefeng Liu, Jianwei Niu, and et al. Dualfed: enjoying both generalization and per- sonalization in federated learning via hierachical represen- tations. In MM, pages 11060–11069, 2024

    2024

  35. [2008]

    Fedccrl: Federated domain generalization with cross-client representation learning

    [Wang and Tang, 2024] Xinpeng Wang and Xiaoying Tang. Fedccrl: Federated domain generalization with cross-client representation learning. arXiv preprint arXiv:2410.11267,

    2024 arXiv

  36. [2011]

    McKeown, and et al

    [Yu et al., 2023] Xinhui Yu, Dan Wang, Martin J. McKeown, and et al. Contrastive-enhanced domain generalization with federated learning. IEEE Transactions on Artificial Intelligence, 5(4):1525–1532,

    2023

  37. [2016]

    Gitfl: Uncertainty-aware real-time asynchronous federated learning using version control

    [Hu et al., 2023] Ming Hu, Zeke Xia, Dengke Yan, and et al. Gitfl: Uncertainty-aware real-time asynchronous federated learning using version control. In RTSS, pages 145–157. IEEE,

    2023

  38. [2017]

    Learn- ing using privileged information for food recognition

    [Meng et al., 2019] Lei Meng, Long Chen, and et al. Learn- ing using privileged information for food recognition. In MM, pages 557–565,

    2019

  39. [2019]

    Improving global generalization and local personalization for federated learning

    [Meng et al., 2024] Lei Meng, Zhuang Qi, Lei Wu, Du Xi- aoyu, and et al. Improving global generalization and local personalization for federated learning. IEEE Transactions on Neural Networks and Learning Systems, 36,

    2024

  40. [2020]

    Model-contrastive federated learning

    [Li et al., 2021] Qinbin Li, Bingsheng He, and Dawn Song. Model-contrastive federated learning. In CVPR, pages 10713–10722,

    2021

  41. [2021]

    A swiss army knife for heterogeneous federated learning: Flexible coupling via trace norm

    [Liao et al., 2024] Tianchi Liao, Lele Fu, Jialong Chen, and et al. A swiss army knife for heterogeneous federated learning: Flexible coupling via trace norm. NeurIPS, 37:139886–139911,

    2024

  42. [2022]

    Ten challenging problems in federated foundation models

    [Fan et al., 2025] Tao Fan, Hanlin Gu, et al. Ten challenging problems in federated foundation models. IEEE Transac- tions on Knowledge and Data Engineering,

    2025

  43. [2023]

    Fair federated learning under domain skew with local consistency and domain diversity

    [Chen et al., 2024] Yuhang Chen, Wenke Huang, and Mang Ye. Fair federated learning under domain skew with local consistency and domain diversity. InCVPR, pages 12077– 12086,

    2024

  44. [2024]

    Federated domain generalization for image recognition via cross-client style transfer

    [Chen et al., 2023] Junming Chen, Meirui Jiang, and et al. Federated domain generalization for image recognition via cross-client style transfer. InWACV, pages 361–370,

    2023

  45. [2025]

    Out-of-distribution generalization of federated learn- ing via implicit invariant relationships

    [Guo et al., 2023] Yaming Guo, Kai Guo, Xiaofeng Cao, and et al. Out-of-distribution generalization of federated learn- ing via implicit invariant relationships. In ICML, pages 11905–11933. PMLR,

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.