Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Towards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A federated prompt-tuning method keeps multi-expert prompts privacy-efficient and achieves a better personalization-generalization trade-off under local differential privacy.

desk verdict Solid empirical paper on federated prompt tuning with real novelty in subspace-decomposed experts, but the logit-level fusion is an unexamined approximation and the privacy claim outruns the DP guarantee. read the letter →

arxiv 2607.21417 v1 pith:WVQ6GD2M submitted 2026-07-23 cs.CV

Yuhua Wang , Xiaodong Li , Yihao Guo , Yuxiang Jia , Qinnan Zhang , Yifan Sun , Hainan Zhang , Yongxin Tong
show 1 more author
Zhiming Zheng
This is my paper
classification cs.CV
keywords federatedprompttuninglocaldifferentialprivacymulti-expertpromptslow-rankdecompositioninstance-awareroutingvision-languagemodelsdataheterogeneitypersonalization-generalizationtrade-off
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 introduces FedSEPT, a federated prompt-tuning method that keeps the benefits of multiple prompt experts while staying cheap under local differential privacy. It claims that a single shared prompt over-smooths the diverse transferable knowledge held by heterogeneous clients, so it replaces it with several expert prompts, each expressed as a low-rank factor times a fixed public basis plus a private residual. Only the small shared factors are privatized and uploaded, so DP noise and communication scale with the factor dimension rather than with the number of full experts. An on-device router then combines the experts per image, using cached text features to fuse at the logit level instead of re-encoding prompts. Under the same privacy budget, the method reports better harmonic-mean accuracy across 11 heterogeneous benchmarks than existing baselines, cutting per-round communication from 64 KiB to 8 KiB.

What carries the argument

Subspace-Decomposed Expert Modeling (SEM) with a fixed public basis: each of M experts is parameterized as P^m_k = A^m_k B0 + R_k, where A^m_k is the shared low-rank factor, B0 is a fixed public basis aligning all clients in a common coordinate system, and R_k is a private residual that never leaves the device. This is paired with Instance-Aware Expert Fusion (IEF), which routes each image through learned weights pi_k(x) and fuses cached per-expert text features at the logit level: z_k,c(x) = sum_m pi_k,m(x) z^m_k,c(x). The machinery confines DP noise and communication to the M L r factor space, enables direct server aggregation, and avoids repeated text-encoder passes.

What would settle it

Compare the fused logits from Eq. (9) against the logits obtained by actually encoding the weighted prompt P_k(x) = (sum_m pi_k,m(x) A^m_k) B0 + R_k on a held-out set, using the same trained experts and router. If the cosine similarity between the two logit vectors drops below a high threshold on any benchmark, the late-fusion approximation is the point of failure; the paper gives no error bound for this step.

Watch

Extended reading notes

Core claim

The central discovery is that multi-expert prompt tuning can be made privacy-efficient by decomposing each expert prompt into a shared low-rank factor A^m_k, a fixed public basis B0, and a private residual R_k. Because the basis is shared and fixed, expert factors live in one coordinate system and can be averaged directly on the server; because only factors are transmitted, DP-SGD perturbs a compact space, avoiding the noise blow-up that would come from privatizing full-dimensional expert prompts. The second piece is instance-aware expert fusion: a client-specific router predicts per-image weights, and the experts' cached class-level text features are combined by weighted logits. The paper r

Load-bearing premise

The method assumes that fusing separately encoded expert logits with per-image weights closely reproduces the logits of the single weighted prompt, so that the text encoder's output is nearly linear in the prompt mixture; if that approximation fails, the instance-aware fusion gains could disappear.

Editorial extensions

If this is right

  • DP noise and communication cost no longer grow with the number of expert prompts or the full prompt dimension; they scale with the M L r factor space, making multi-expert collaboration practical on bandwidth-limited clients.
  • The same privacy budget buys a better personalization-generalization trade-off: harmonic-mean accuracy improves on fine-grained, label-skew, and domain-shift benchmarks, e.g., OxfordPets HM 90.24 vs 80.67 at epsilon=1.
  • Instance-aware logit-level fusion with cached text features keeps inference cheap: 2.35 ms latency and 35.1 GFLOPs, lower than prompt-level fusion, while retaining comparable memory during training.
  • Local DP suppresses both membership inference and gradient inversion: final MIA AUC drops to 0.45-0.50 under update-only attacks and to 0.55-0.58 under update-and-query attacks, and CLIP cosine similarity for gradient inversion becomes negative at epsilon=1.
  • A single framework covers pathological label skew, practical Dirichlet label skew, and compound domain+label skew, suggesting the expert decomposition is a general remedy for over-smoothed shared prompts.

Reading between the lines

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

  • Because the public basis is fixed and data-independent, the method implicitly assumes the transferable directions of every task lie in its column space; the paper does not analyze this, so a natural check is to vary the basis initialization and rank and measure performance.
  • The same decomposition-plus-cached-logit recipe is not tied to text prompts: it could apply to other frozen encoders and other parameter-efficient modules (e.g., adapters), wherever per-module outputs can be cached and fused per instance.
  • Only the shared factors are protected by the stated local-DP guarantee; the private residual and router are never uploaded. In deployments where attackers can query the final personalized model, the paper's own query-based MIA results imply these unshared states need extra protection.
  • The diversity regularizer and routing weights may double as a diagnostic: inspecting per-class or per-domain routing distributions could reveal which experts capture which latent semantic or style clusters, giving a data-driven view of client heterogeneity.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. FedSEPT proposes a federated prompt-tuning method for heterogeneous clients under local differential privacy. It parameterizes each of M prompt experts as A_m B0 + R_k, with a fixed public basis B0 and a client-private residual R_k; only the low-rank factors A_m are communicated and aggregated, reducing communication cost and the dimension of DP perturbation. An instance-aware router computes soft expert weights from image features; rather than constructing an input-dependent prompt, the method caches per-expert text features and fuses logits (Eq. 9). Experiments on 11 benchmarks report In-Client, Cross-Client, and harmonic-mean accuracy against six baselines, together with ablations, efficiency measurements, and membership-inference/gradient-inversion attack evaluations. The central claim is that under the same privacy budget, FedSEPT achieves a better local-adaptation/global-generalization trade-off than existing baselines.

Significance. If the empirical claims hold, FedSEPT is a useful contribution to privacy-preserving federated prompt tuning. The subspace-decomposed expert parameterization is a practical way to scale to multiple experts under local DP, and the reported communication reduction (64 to 8 KiB per round) is substantial. The paper ships code, uses a standard DP-SGD privacy proof, and evaluates both utility and privacy attacks, which is a strength. Notable reported gains include OxfordPets HM 90.24 vs 80.67 for DP-FPL at ε=1, and consistent HM improvements on several other datasets. However, the central mechanism of 'instance-aware expert fusion' rests on an unvalidated logit-level linearity approximation, the empirical headline differences are not accompanied by uncertainty estimates or significance tests, and the 'privacy-preserving' label is stronger than the reported MIA results support. These issues are load-bearing for the paper's main claims and require revision.

major comments (3)
  1. The paper motivates IEF as composing experts into an input-dependent prompt P_k(x) = (Σ_m π_m A_m) B0 + R_k, but Eq. (9) actually fuses cached per-expert logits: z_{k,c}(x) = Σ_m π_{k,m}(x) ⟨v(x), t^m_{k,c}⟩/τ. This is exactly equivalent to the logit of the weighted prompt only if the CLIP text encoder is linear in the prompt parameters, which is false for a transformer. No error bound or empirical validation of this approximation is given. Table 5 compares FedSEPT-PF (prompt-level fusion) only on latency, FLOPs, and memory, not on accuracy. As written, the method is a logit-level ensemble, not the advertised instance-wise prompt composition, and the 'multi-expert collaboration' claim in the conclusion goes beyond what is demonstrated. Please add an accuracy comparison between Eq. (9) and true prompt-level fusion (or an approximation-error measurement) on at least one representative benc
  2. The main empirical results are reported as point estimates only: the text states 'mean client performance over the last 5 rounds', but no standard deviations, confidence intervals, or number of seeds are given. Several headline differences are small, e.g., Food101 HM 80.75 vs 80.67 for DP-FPL in Table 1, and DomainNet HM 67.07 vs 66.79 at β=0.1 in Table 3. Without multiple seeds and significance testing, the central claim of a better trade-off is not robustly supported. Please report mean±std over at least three seeds for the main tables, and state whether the reported gains are statistically significant for the close comparisons.
  3. The update-and-query MIA reaches final AUC 0.5521–0.5790 for ε∈{0.25,0.5,0.75,1.0}, down from 0.9471 without DP but still substantially above random guessing. Local DP protects only the uploaded shared factors; it does not protect the outputs of the final personalized model, which the query-based attacker exploits. The unqualified 'privacy-preserving' phrasing in the title, abstract, and conclusion is therefore too strong. Please add an explicit statement of the threat model (local DP covers server-visible updates only), report the query-attack residual leakage as a known limitation, and qualify the privacy claim accordingly.
minor comments (5)
  1. The 'Single' ablation is described as removing both SEM and IEF, but the exact parameterization is not fully specified. State whether it is a single full-dimensional prompt of length L, and clarify how it is aggregated under DP.
  2. Algorithm 1 says to compute fused predictions using Eq. (9), but the caching of per-expert text features is not explicit. Clarify the timing: per-expert text features must be recomputed after each prompt update and then cached for reuse across the minibatch.
  3. The DP-SGD formulation in Eq. (4) writes the noise as added to the minibatch average. Standard DP-SGD often clips per-example gradients and adds noise to the sum before dividing by batch size; the two are equivalent up to a rescaling of the noise multiplier, but this should be stated to avoid confusion.
  4. The line plots for seven methods are dense and small at the current size. Consider larger panels, markers for individual methods, or shaded error regions if multiple seeds are added.
  5. The DP-FPL baseline applies QR-based re-factorization after each local epoch. Please state explicitly how this operation interacts with the privacy accounting (e.g., as a deterministic post-processing step) so that the comparison is not open to a privacy-accounting objection.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: performance claims are benchmarks against external baselines; the DP guarantee is standard analysis; logit-level fusion is an untested approximation, not a definitional identity.

full rationale

The paper's central claims are empirical comparisons against externally published baselines (PromptFL, FedPGP, FedOTP, FedPHA, pFedMoAP, DP-FPL) on 11 benchmarks; no target result is used to define a fitted parameter or to construct a prediction. SEM's factorization and expert-wise aggregation are design choices motivated by low-rank adaptation and coordinate alignment, with a standard argument about non-uniqueness of low-rank factorizations. The privacy guarantee is the standard DP-SGD moments accountant applied to clipped and noised factor updates, citing Abadi et al. as external analysis; it is not imported from the authors' own prior uniqueness results. The only self-reference by overlapping authors ([37]) is not load-bearing for the main derivation. The strongest concern is Eq. (9): logit-level fusion is not equal to the true prompt-level composition unless the text encoder is linear in prompt space, and no error bound or accuracy comparison versus prompt-level fusion is given. However, the paper explicitly presents this as a late-fusion strategy for efficiency, not as an identity, and the reported gains are empirical benchmark results that stand or fall independently of the approximation. This is a correctness risk, not circularity: no fitted input is renamed as a prediction, no self-citation chain forces the design, and no known pattern is merely relabeled. Therefore a non-finding is appropriate.

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

The main free choices are regularization weights, expert count, and subspace rank, chosen by analysis on the same benchmark families. The method rests on the low-rank adaptation assumption and on treating logit-level fusion as equivalent to prompt-level fusion; neither is formally established.

free parameters (4)
  • lambda_div = 10
    Selected via grid search in Fig. 4(a) on OxfordPets/OfficeHome; no held-out split described.
  • lambda_gate = 10
    Selected via grid search in Fig. 4(a); no held-out split described.
  • number of experts M = 4
    Chosen from Fig. 4(b) as robust default; larger M yields no consistent gains.
  • subspace rank r = 16
    Chosen from Fig. 4(c) as a balance between representation capacity and parameter efficiency.
assumptions (4)
  • domain assumption Adaptation updates of frozen pretrained models are effectively low-rank/low-dimensional
    Intro and §4.1 rely on this to justify SEM; cites [2,17] but it is not formally established for this setting.
  • domain assumption A fixed public basis B0 aligns expert factors so that server-side averaging is a valid aggregation
    §4.1.1: low-rank factorizations are non-unique; fixing B0 avoids coordinate mismatch, but assumes the fixed random basis spans a useful subspace for all clients.
  • domain assumption Logit-level weighted fusion with cached text features faithfully approximates prompt-level fusion
    §4.2.2 Eq. (9) replaces a weighted prompt combination with a weighted combination of per-expert logits; no bound on the approximation error is given.
  • standard math Standard DP-SGD moments-accountant composition applies to this per-step subsampled Gaussian mechanism
    Theorem 4.1 invokes [1]; this is standard when clipping/noise are applied to the shared factors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach." pith.science (2026). https://pith.science/paper/WVQ6GD2M

@misc{pith2026260721417,
  author       = {Pith},
  title        = {Pith review of: Towards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WVQ6GD2M}},
  note         = {Machine review of arXiv:2607.21417}
}
read the original abstract

Federated prompt tuning (FPT) enables collaborative adaptation of vision--language models (VLMs) using lightweight prompts. Existing methods often address heterogeneity and privacy through a split-prompt design under local differential privacy (DP), combining a shared prompt for global transfer with private prompts for local adaptation. However, a single shared prompt may over-smooth diverse transferable knowledge, weakening the balance between personalization and generalization. Multi-expert prompts (MEPs) can better capture this diversity, but enlarge the communicated space, increasing DP noise and communication cost while making robust expert composition more difficult. We propose FedSEPT, a privacy-preserving Fed}erated Subspace-decomposed Expert Prompt Tuning. Specifically, we employ Subspace-decomposed Expert Modeling (SEM) to parameterize multiple prompt experts with shared low-rank factors, a fixed public basis, and private residuals, thereby confining communication and DP perturbation to a compact factor space while enabling direct server aggregation in a common coordinate system. We further design Instance-aware Expert Fusion (IEF), which adaptively combines semantically complementary experts via on-device routing and performs efficient logit-level fusion using cached expert-specific text features. Extensive experiments on 11 heterogeneous benchmarks show that, under the same privacy constraints, FedSEPT achieves a better trade-off between local adaptation and global generalization than strong baselines.

Figures

Figures reproduced from arXiv: 2607.21417 by the authors.

Figure 1
Figure 1. Motivation. (a) Privacy-Preserving FPT suffers from [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of FedSEPT. ① Each client constructs 𝑀 prompt experts as P 𝑚 𝑘 = A 𝑚 𝑘 B0 + R𝑘 , where DP-SGD is applied only to the shared factors A 𝑚 𝑘 , the public basis B0 remains fixed, and the private residual R𝑘 stays local. ② A local router predicts instance-aware expert weights from image features. ③ Expert text features are cached and combined through logit-level fusion. ④ Clients upload only privatized expert fa… view at source ↗
Figure 3
Figure 3. Effect of the number of shots. 5.2 Performance Comparison Pathological Label Skew and Data Scarcity. As shown in Tab. 1, FedSEPT achieves the best HM on all five datasets under pathologi￾cal label skew. Its gains are particularly pronounced on Food101, DTD, and OxfordPets, where stronger Cross-Client accuracy than highly personalized methods such as FedOTP and FedPHA leads to substantially higher HM while retaining … view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Hyperparameter analysis. Practical Label Skew. Tab. 2 shows that FedSEPT remains consis￾tently effective in larger-scale and more realistic federated settings. While PromptFL and FedPGP are competitive in some Cross-Client cases, they are generally weaker in In-Client …
Figure 5
Figure 5. Figure 5: MIAs under two threat models on OxfordPets. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 2 linked inside Pith

  1. [1]

    Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. 2016. Deep learning with differential privacy. In Proceedings of the 2016 ACM SIGSAC conference on computer and communications security. 308–318

    2016

  2. [2]

    Armen Aghajanyan, Sonal Gupta, and Luke Zettlemoyer. 2021. Intrinsic Di- mensionality Explains the Effectiveness of Language Model Fine-Tuning. In Proceedings of the 59th Annual Meeting of the Association for Computational Lin- guistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Association for Comp...

    2021

  3. [3]

    Li Bai, Haibo Hu, Qingqing Ye, Haoyang Li, Leixia Wang, and Jianliang Xu. 2024. Membership Inference Attacks and Defenses in Federated Learning: A Survey. ACM Comput. Surv.57, 4, Article 89 (Dec. 2024), 35 pages

    2024

  4. [4]

    Sikai Bai, Jie Zhang, Song Guo, Shuaicheng Li, Jingcai Guo, Jun Hou, Tao Han, and Xiaocheng Lu. 2024. DiPrompT: Disentangled Prompt Tuning for Multiple Latent Domain Generalization in Federated Learning. In2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 27274–27283

    2024

  5. [5]

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. 2014. Food-101–mining discriminative components with random forests. InEuropean conference on com- puter vision. Springer, 446–461

    2014

  6. [6]

    Mircea Cimpoi, Subhransu Maji, Iasonas Kokkinos, Sammy Mohamed, and An- drea Vedaldi. 2014. Describing Textures in the Wild. InProceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR ’14). IEEE Computer Society, USA, 3606–3613

    2014

  7. [7]

    Tianyu Cui, Hongxia Li, Jingya Wang, and Ye Shi. 2024. Harmonizing general- ization and personalization in federated prompt learning. InProceedings of the 41st International Conference on Machine Learning(Vienna, Austria)(ICML’24). JMLR.org, Article 381, 16 pages

    2024

  8. [8]

    Cynthia Dwork. 2006. Differential privacy. InInternational colloquium on au- tomata, languages, and programming. Springer, 1–12

    2006

Show all 49 references
  1. [9]

    Chengying Fang, Wenke Huang, Guancheng Wan, Yihao Yang, and Mang Ye. 2025. FedPHA: Federated Prompt Learning for Heterogeneous Client Adaptation. In Proceedings of the 42nd International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 267). PMLR,...

    2025

  2. [10]

    William Fedus, Barret Zoph, and Noam Shazeer. 2022. Switch transformers: scaling to trillion parameter models with simple and efficient sparsity.J. Mach. Learn. Res.23, 1, Article 120 (Jan. 2022), 39 pages

    2022

  3. [11]

    Li Fei-Fei, Rob Fergus, and Pietro Perona. 2004. Learning Generative Visual Models from Few Training Examples: An Incremental Bayesian Approach Tested on 101 Object Categories. InProceedings of the 2004 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW ’04)...

    2004

  4. [12]

    Jonas Geiping, Hartmut Bauermeister, Hannah Dröge, and Michael Moeller. 2020. Inverting Gradients: How Easy Is It to Break Privacy in Federated Learning?. In Advances in Neural Information Processing Systems, Vol. 33. 16937–16947

    2020

  5. [13]

    Boqing Gong, Yuan Shi, Fei Sha, and Kristen Grauman. 2012. Geodesic flow kernel for unsupervised domain adaptation. In2012 IEEE conference on computer vision and pattern recognition. IEEE, 2066–2073

    2012

  6. [14]

    Tao Guo, Song Guo, and Junxiao Wang. 2023. Pfedprompt: Learning personalized prompt for vision-language models in federated learning. InProceedings of the ACM Web Conference 2023. 1364–1374

    2023

  7. [15]

    Tao Guo, Song Guo, Junxiao Wang, Xueyang Tang, and Wenchao Xu. 2023. Promptfl: Let federated participants cooperatively learn prompts instead of models–federated learning in age of foundation model.IEEE Transactions on Mobile Computing23, 5 (2023), 5179–5194

    2023

  8. [16]

    Sizai Hou, Songze Li, and Baturalp Buyukates. 2025. Privacy-preserving Prompt Personalization in Federated Learning for Multimodal Large Language Models. arXiv:2505.22447 [cs.CR] https://arxiv.org/abs/2505.22447

    2025 arXiv

  9. [17]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-Rank Adaptation of Large Language Models. InInternational Conference on Learning Representations

    2022

  10. [18]

    Yongqi Huang, Peng Ye, Xiaoshui Huang, Sheng Li, Tao Chen, Tong He, and Wanli Ouyang. 2023. Experts weights averaging: A new general training scheme for vision transformers.arXiv preprint arXiv:2308.06093(2023)

    2023 arXiv

  11. [19]

    Peter Kairouz and H Brendan McMahan. 2021. Advances and open problems in federated learning.Foundations and trends in machine learning14, 1-2 (2021), 1–210

    2021

  12. [20]

    Alex Krizhevsky, Geoffrey Hinton, et al. 2009. Learning multiple layers of features from tiny images. (2009)

    2009

  13. [21]

    Seanie Lee, Sangwoo Park, Dong Bok Lee, Dominik Wagner, Haebin Seong, Tobias Bocklet, Juho Lee, and Sung Ju Hwang. 2025. FedSVD: Adaptive Orthogonal- ization for Private Federated Learning with LoRA. InThe Thirty-ninth Annual Conference on Neural Information Processing Systems

    2025

  14. [22]

    Hongxia Li, Wei Huang, Jingya Wang, and Ye Shi. 2024. Global and local prompts cooperation via optimal transport for federated learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 12151–12161

    2024

  15. [23]

    Jun Luo, Chen Chen, and Shandong Wu. 2025. Mixture of Experts Made Personal- ized: Federated Prompt Learning for Vision-Language Models. InThe Thirteenth International Conference on Learning Representations

    2025

  16. [24]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Agüera y Arcas. 2017. Communication-Efficient Learning of Deep Net- works from Decentralized Data. InProceedings of the International Conference on Artificial Intelligence and Statistics (AISTATS)

    2017

  17. [25]

    Milad Nasr, Reza Shokri, and Amir Houmansadr. 2019. Comprehensive privacy analysis of deep learning: Passive and active white-box inference attacks against centralized and federated learning. In2019 IEEE symposium on security and privacy (SP). IEEE, 739–753

    2019

  18. [26]

    Quan Minh Nguyen, Min-Seon Kim, Hoang M Ngo, Trong Nghia Hoang, Hyuk- Yoon Kwon, and My T Thai. 2026. Leveraging Soft Prompts for Privacy Attacks in Federated Prompt Tuning.arXiv preprint arXiv:2601.06641(2026)

    2026

  19. [27]

    Maria-Elena Nilsback and Andrew Zisserman. 2008. Automated Flower Clas- sification over a Large Number of Classes. In2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing. 722–729

    2008

  20. [28]

    Omkar M Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. 2012. Cats and dogs. In2012 IEEE Conference on Computer Vision and Pattern Recognition. 3498–3505

    2012

  21. [29]

    Xingchao Peng, Qinxun Bai, Xide Xia, Zijun Huang, Kate Saenko, and Bo Wang

  22. [30]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models From Natural Language Supervision. InProceedings ...

    2021

  23. [31]

    Kate Saenko, Brian Kulis, Mario Fritz, and Trevor Darrell. 2010. Adapting visual category models to new domains. InEuropean conference on computer vision. Springer, 213–226

    2010

  24. [32]

    Roy-Chowdhury, Srikanth V

    Hasin Us Sami, Swapneel Sen, Amit K. Roy-Chowdhury, Srikanth V. Krishna- murthy, and Basak Guler. 2025. Gradient Inversion Attacks on Parameter-Efficient Fine-Tuning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 10224–10234

    2025

  25. [33]

    Noam Shazeer, Azalia Mirhoseini, Krzysztof Maziarz, Andy Davis, Quoc Le, Geoffrey Hinton, and Jeff Dean. 2017. Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer. InInternational Conference on Learning Representations

    2017

  26. [34]

    Linh Tran, Wei Sun, Stacy Patterson, and Ana Milanova. 2025. Privacy-Preserving Personalized Federated Prompt Learning for Multimodal Large Language Models. InThe Thirteenth International Conference on Learning Representations

    2025

  27. [35]

    Hemanth Venkateswara, Jose Eusebio, Shayok Chakraborty, and Sethuraman Panchanathan. 2017. Deep hashing network for unsupervised domain adaptation. InProceedings of the IEEE conference on computer vision and pattern recognition. 5018–5027

    2017

  28. [36]

    Chendi Wang, Yuqing Zhu, Weijie J Su, and Yu-Xiang Wang. 2024. Neural Collapse meets Differential Privacy: Curious behaviors of NoisyGD with Near-Perfect Representation Learning. InProceedings of the 41st International Conference on Machine Learning (Proceedings of Machine Lea...

    2024

  29. [37]

    Yuhua Wang, Qinnan Zhang, Xiaodong Li, Huan Zhang, Yifan Sun, Wangjie Qiu, Hainan Zhang, Yongxin Tong, and Zhiming Zheng. 2026. Taming Noise- Induced Prototype Degradation for Privacy-Preserving Personalized Federated Fine-Tuning. InProceedings of the IEEE/CVF Conference on Co...

    2026

  30. [38]

    Vincent Poor

    Kang Wei, Jun Li, Ming Ding, Chuan Ma, Hang Su, Bo Zhang, and H. Vincent Poor. 2022. User-Level Privacy-Preserving Federated Learning: Analysis and Performance Optimization.IEEE Transactions on Mobile Computing21, 9 (2022), 3388–3401

    2022

  31. [39]

    Fu-En Yang, Chien-Yi Wang, and Yu-Chiang Frank Wang. 2023. Efficient model personalization in federated learning via client-specific prompt generation. In Proceedings of the IEEE/CVF International Conference on Computer Vision. 19159– 19168

    2023

  32. [40]

    Yihao Yang, Wenke Huang, Guancheng Wan, Bin Yang, and Mang Ye. 2025. Fed- erated Disentangled Tuning with Textual Prior Decoupling and Visual Dynamic Adaptation. InForty-second International Conference on Machine Learning

    2025

  33. [41]

    Hantao Yao, Rui Zhang, and Changsheng Xu. 2023. Visual-language prompt tun- ing with knowledge-guided context optimization. InProceedings of the IEEE/CVF MM ’26, November 10–14, 2026, Rio de Janeiro, Brazil Wang et al. conference on computer vision and pattern recognition. 6757–6767

    2023

  34. [42]

    Yuen, and Dacheng Tao

    Mang Ye, Xiuwen Fang, Bo Du, Pong C. Yuen, and Dacheng Tao. 2023. Hetero- geneous Federated Learning: State-of-the-art and Research Challenges.ACM Comput. Surv.56, 3, Article 79 (Oct. 2023), 44 pages

    2023

  35. [43]

    Haodong Zhao, Wei Du, Fangqi Li, Peixuan Li, and Gongshen Liu. 2023. Fed- prompt: Communication-efficient and privacy-preserving prompt tuning in fed- erated learning. InICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5

    2023

  36. [44]

    Rajapakse

    Yubin Zheng, Pak Hei Yeung, Jing Xia, Tianjie Ju, Peng Tang, Weidong Qiu, and Jagath C. Rajapakse. 2025. FedDEAP: Adaptive Dual-Prompt Tuning for Multi-Domain Federated Learning. InProceedings of the 33rd ACM International Conference on Multimedia(Dublin, Ireland)(MM ’25). Ass...

    2025

  37. [45]

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. 2022. Conditional prompt learning for vision-language models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 16816–16825

    2022

  38. [46]

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. 2022. Learning to prompt for vision-language models.International journal of computer vision 130, 9 (2022), 2337–2348

    2022

  39. [47]

    Yingxue Zhou, Steven Wu, and Arindam Banerjee. 2021. Bypassing the Ambient Dimension: Private SGD with Gradient Subspace Identification. InInternational Conference on Learning Representations

    2021

  40. [48]

    Ligeng Zhu, Zhijian Liu, and Song Han. 2019. Deep Leakage from Gradients. In Advances in Neural Information Processing Systems, Vol. 32. 14774–14784. Towards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach MM ’26, Nove...

    2019

  41. [2019]

    InProceedings of the IEEE/CVF international conference on computer vision

    Moment matching for multi-source domain adaptation. InProceedings of the IEEE/CVF international conference on computer vision. 1406–1415

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.