Pith. sign in

REVIEW 3 major objections 5 minor 51 references

Lethe: How Hard Is It to Forget? A Benchmark for Federated Unlearning in Medical Imaging

T0 review · 3 major / 5 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 Forgetting difficulty, not method choice, decides whether federated unlearning methods can be told apart in medical imaging.

desk verdict A serious benchmark with a real meta-finding, but the difficulty variable is measured wrong (F0 instead of the erasable gap) and BFU degenerates to GA on several datasets; revise before relying on the rankings. read the letter →

arxiv 2608.01094 v1 pith:4MIWJO7N submitted 2026-08-02 cs.CV

Shengchao Chen , Ting Shu This is my paper · ORCID
classification cs.CV
keywords federatedunlearningmachinemedicalimagingbenchmarkmembershipinferenceforgettingdifficultycross-sitegeneralizationrighttobeforgotten
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

Lethe is a benchmark for federated unlearning in medical imaging that spans eight task families, sixteen datasets, twelve unlearning methods, and three forgetting granularities. Its central claim is that the difficulty of the forgetting request, not the choice of unlearning method, determines whether methods can be told apart. On easy client-level removals, all methods that preserve utility cluster indistinguishably close to a from-scratch retrained gold standard, and only class-level or sole-class removals widen the spread. On clinical tasks that generalize across sites, forgetting a client barely changes the task metric, so the signal an unlearner must erase is residual membership information rather than task contribution. The paper concludes that federated unlearning evaluation should be difficulty-aware and target membership.

What carries the argument

The load-bearing object is the Lethe protocol itself, a shared evaluation where every unlearned model is scored against a gold standard retrained from scratch on the retained data, and where three request granularities (client, class, sample) and task families with different cross-site generalization are varied systematically. Inside that protocol, the operative quantity is $F_0$, the pre-unlearn model's accuracy on the forget set, which the paper uses as a difficulty meter: $F_0$ falls from about 0.49 on BloodMNIST to 0.05 on PathMNIST, and method separability tracks this ordering. The protocol also substitutes a retrain-calibrated per-sample membership attack for population-level attacks, because membership is the signal that survives when task transfer makes forget accuracy vacuous.

What would settle it

On a single dataset, hold architecture, method set, and task transfer fixed and vary $F_0$ by changing which classes the forget client holds exclusively; if the spread in retain accuracy among the gentle methods does not increase with $F_0$, the claim that difficulty governs separability fails.

Watch

Extended reading notes

Core claim

Lethe's central discovery is empirical: what separates federated unlearning methods is the hardness of the forgetting request, read from the pre-unlearn forget accuracy $F_0$, not the algorithm itself. Easy requests, like a departing client whose classes are also held elsewhere, leave utility-preserving methods statistically indistinguishable from each other and from the retrained gold standard, while hard requests—class-level removal, a sole-holder client leaving, hard 3D classification—produce measurable method spread. A second discovery is that many clinical tasks generalize across sites: the gold retrain, which never saw the forget client, still performs on its data within a few points of every method, making forget accuracy a near-vacuous metric. Under a retrain-calibrated per-sample membership attack, residual membership remains detectable for several gentle methods, which is why the paper identifies membership, not task accuracy, as the real erasure target.

Load-bearing premise

The argument assumes that the pre-unlearn forget accuracy $F_0$ measures request difficulty independently of the compared methods and of how much the task transfers across sites, so that method-spread can be attributed to difficulty rather than to architecture, label skew, or task generalization.

Editorial extensions

If this is right

  • Benchmarks should stress class-level and sole-class removal, because easy client-level requests cannot separate methods.
  • Privacy evaluation should report retrain-calibrated per-sample membership-inference scores, since forget accuracy can be matched while residual membership remains.
  • An unlearning system should estimate request difficulty from class support, sole-provider status, and site skew, and escalate to full retraining only on hard requests.
  • Client-level method rankings obtained on natural-image, single-heterogeneity benchmarks should not be assumed to transfer to medical imaging.
  • Forgetting should be treated as a design property: site-scoped or class-scoped adapters that can be dropped or retrained are a more promising direction than post-hoc weight surgery.

Reading between the lines

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

  • One extension the paper does not run: use representation-level difficulty, such as the distance between forget-class features and the retained manifold, to see whether methods separate even on easy requests.
  • The task-transfer result likely generalizes beyond medicine: any federated deployment with redundant sites or strong server-side generalization can hide residual membership behind stable task metrics, so calibrated per-sample attacks should be standard there too.
  • If $F_0$ truly governs separability, a cheap pre-unlearn measurement could be used at request time to decide whether to skip unlearning, apply a gentle update, or pay for retraining.
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

3 major / 5 minor

Summary. The paper introduces Lethe, a benchmark for federated unlearning in medical imaging, spanning eight task families, sixteen datasets, twelve unlearning methods, three forgetting granularities (client, class, sample), and a seven-metric panel evaluated against a retrained gold standard. The central claimed finding is that the difficulty of the forgetting request, rather than the choice of unlearning method, determines whether methods can be distinguished: on easy client-level removals, utility-preserving methods are statistically indistinguishable, while hard class-level and sole-class removals separate methods. A second claim is that on clinical tasks that generalize across sites, client forgetting barely changes task performance, leaving residual membership as the primary erasable signal. The authors validate this with a Friedman-Nemenyi ranking, several robustness checks (heterogeneity, scale, architecture, sample-level forgetting), per-sample membership inference attacks, and case studies, and they release code and data generators.

Significance. If the central claim holds, the paper makes a useful contribution to federated unlearning evaluation: it would show that the field's client-level benchmarks are non-discriminative, that hard requests are necessary for method comparison, and that membership leakage rather than task-metric change is the relevant target on many medical tasks. The benchmark itself is large and well-scoped, with a shared protocol, a gold-standard retrain baseline, per-sample MIA, and cost/durability metrics, and the authors provide code and data-generation scripts, which supports reproducibility. The finding that 'gentle' methods are often statistically indistinguishable is supported by the reported Friedman-Nemenyi analysis, and the qualitatively consistent pattern across many tables is a genuine strength. However, the headline causal claim about difficulty is under-supported because the difficulty proxy used (F0) appears internally inconsistent with the paper's own segmentation results, and the evidence is correlational rather than manipulative.

major comments (3)
  1. [§4.1, Fig. 2(b), §4.2, Table 6] The paper operationalizes 'forgetting difficulty' as the pre-unlearn forget accuracy F0 and uses it to conclude that difficulty governs separability (Sec. 4.1). This proxy conflates two regimes: high F0 with low gold retrain forget accuracy F* (a real erasable signal) versus high F0 with high F* (task transfer, little to erase). The paper's own segmentation results contradict the stated F0 rule: Table 6 reports pre-unlearn forget Dice F0 of 0.570–0.890, which under the F0-as-difficulty reading would make segmentation the hardest setting and should separate methods, yet the text reports that 'the gold retrain ... still segments its data within a few percentage points of every method' and that 'the forget metric is nearly vacuous' (Sec. 4.2). The quantity that actually tracks the observed separability is the erasable gap D = F0 − F*, not F0. The authors should retest the headline claim using D and should report the correlation of F0 and D with a quantitative measure of method spread (e.g., variance of retain accuracy or the Friedman test statistic). Without this, the recommendation to route requests by hardness (Sec. 4) is not supported.
  2. [§3, Table 3; App. B.2, Fig. 6] Several reported results show exact numerical duplication between methods from different mechanism families: BFU exactly matches GA on OrganAMNIST, PathMNIST, and OCTMNIST (Tables 15–17, F/R columns identical to GA), and Table 7 shows many methods identical to three decimals in the open-ended VQA block. Exact duplication of a Bayesian method (BFU) by gradient ascent strongly suggests an implementation bug or that BFU degenerates to GA in those settings. This is load-bearing for the benchmark's credibility as a method zoo: the Friedman–Nemenyi ranking in App. B.2 treats these as distinct methods, and the conclusion that 'only the gradient-ascent methods fall significantly below' is confounded if some non-ascent methods are actually running ascent. The authors should investigate and clarify these exact matches, or remove the affected columns/rows with an explanation.
  3. [§4, 'Where the Field Should Go Next'; App. B.3, Table 18] The causal phrasing 'difficulty governs separability' is not justified by the evidence presented. The paper shows correlations across datasets that differ simultaneously in F0, label skew, architecture, and cross-site generalization, but it never manipulates difficulty within a fixed task and dataset while holding other factors constant. The sole-class protocol in App. B.3 (Table 18) is the natural experiment, but it only reports retain accuracy for a subset of gentle methods and does not report F0, F*, or a separability measure for the same dataset under client-level versus sole-class requests. A concrete test would be to vary the number of holders of a class (or the F0–F* gap) within one dataset and show that method spread tracks that variation. Absent such a test, 'difficulty' remains a post-hoc label for the observed pattern rather than a demonstrated causal variable.
minor comments (5)
  1. [Fig. 2(b)] The y-axis label reads 'gold forget acc', but F0 is the pre-unlearn model's forget accuracy, not the gold retrain's; please relabel to 'pre-unlearn forget accuracy F0'.
  2. [Table 3 caption] The caption says 'Green marks the two retain values nearest gold', but the table is rendered in monochrome in the text; please ensure the color coding is visible or replace with bold/asterisks.
  3. [Eq. (4) and following text] The definition R_k = A(M_u^(k), D_f) is the recovered forget accuracy after k fine-tuning epochs, but the text says 'a durable erasure keeps R_k near M*'; since the gold retrain's forget accuracy is a scalar, please state the comparison target explicitly (e.g., 'near the gold retrain's forget accuracy').
  4. [Table 11] The speedup table reports negative or zero speedup for FedEraser (0.6–1.1×), which is mentioned in the text, but the caption does not note that values below 1 indicate slower-than-retrain; please add a clarifying note.
  5. [§4.3, Table 7] The open-ended VQA block reports many identical values (e.g., 0.296±0.002 for several methods), which may be due to rounding or the reported precision; if so, please state the rounding convention in the caption.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central finding is an empirical benchmark comparison against an external retrained gold standard, not a fitted or self-referential derivation.

full rationale

No circularity found. The central claim—that request difficulty, not method choice, governs how strongly unlearning methods separate—is reported as an empirical correlation across many independently constructed task families. The difficulty variable F0 is measured on the pre-unlearn model (Full model in Table 3 and Fig. 2b), while method spread is measured post-unlearn against a retrained-from-scratch gold standard M*, so the two quantities are not the same by construction. No free parameter is fitted to the target finding, and the gold retrain provides an external anchor: every method is scored relative to M*, not to a fitted target. The F0-based difficulty operationalization is open to a validity concern—Section 4.2 itself calls the forget metric 'nearly vacuous' when cross-site generalization makes the gold forget accuracy close to F0, suggesting that the erasable gap F0−F* rather than F0 alone may be the operative quantity—but that is a measurement-validity or correctness issue, not a circular derivation. The paper does not define method spread in terms of F0, nor does it fit F0 to the observed spread; the correlation is presented as a finding, and the skeptic's alternative variable is a competing explanation rather than a demonstration that the claim reduces to its input by definition. The authors' self-citations ([1], [10]–[16], [47]) appear in background context and future-work suggestions (e.g., site-scoped adapters) and are not load-bearing for the empirical benchmark results. The benchmark is self-contained against external gold retraining and standard public datasets, so the honest finding is no significant circularity.

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

The paper's central finding is empirical and does not depend on fitted constants. It does depend on assumptions about the gold standard, the representativeness of the method zoo, the validity of membership-inference calibration, and the realism of synthetic Dirichlet partitions. None of these are proven, and the BFU/GA duplication suggests at least one method implementation may be unreliable.

assumptions (4)
  • domain assumption FedAvg-trained models and gold-standard retrains are the correct reference for federated unlearning.
    The benchmark defines unlearning as approximation of retraining from scratch on retained data (Eq. 1 to 3). This is the standard legal interpretation but is not proven to be the only or best target.
  • domain assumption The twelve methods are representative implementations of their mechanism families with hyperparameters consistent with original papers.
    The conclusion 'difficulty, not method' assumes the method zoo fairly represents the mechanism space; failures or implementation bugs, such as BFU duplicating GA in Tables 3 and 14 to 17, would bias the comparison.
  • domain assumption Per-sample membership inference calibrated on retrain is a valid measure of residual privacy.
    Privacy conclusions rely on the calibrated LiRA and U-LiRA attacks; the normality assumptions in Eq. 9 are standard but unverified here.
  • domain assumption Dirichlet label-skew partitions on datasets without natural client splits approximate real cross-silo hospital heterogeneity.
    Most classification datasets are artificially partitioned across 20 clients with alpha equal to 0.1; the findings are robust to alpha variations, but synthetic partitions may still not capture real scanner and protocol heterogeneity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Lethe: How Hard Is It to Forget? A Benchmark for Federated Unlearning in Medical Imaging." pith.science (2026). https://pith.science/paper/4MIWJO7N

@misc{pith2026260801094,
  author       = {Pith},
  title        = {Pith review of: Lethe: How Hard Is It to Forget? A Benchmark for Federated Unlearning in Medical Imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4MIWJO7N}},
  note         = {Machine review of arXiv:2608.01094}
}
read the original abstract

Federated learning enables medical-imaging models to be trained across hospitals, and privacy law, most explicitly the GDPR ``right to be forgotten'', turns removing a hospital's, a class's, or a patient's influence from such a model into a federated unlearning problem. This need is most acute in medicine, where patients withdraw consent and hospitals leave collaborations. Yet nearly all unlearning evidence comes from natural images, whose heterogeneity and task structure differ sharply from clinical data, so it is unclear whether existing methods transfer, and no shared protocol covers clinical data. We present Lethe, a benchmark for federated unlearning in medical imaging. It evaluates twelve methods across eight task families, from classification and segmentation to denoising, cross-modality synthesis, and vision-language question answering, at three forgetting granularities and against a retrained gold standard on utility, privacy, and cost. The central result is that what separates methods is the difficulty of the forgetting request, not the method itself. The easy removals that dominate the literature leave the methods that preserve utility indistinguishable, while only hard ones separate them. More striking, on the many medical tasks that generalize across sites, forgetting a client barely changes task performance, leaving residual membership as the signal that must be erased.

Figures

Figures reproduced from arXiv: 2608.01094 by the authors.

Figure 1
Figure 1. Overview of Lethe. A federated model 𝑀0 receives a forgetting request 𝑢 at client, class, or sample granularity, and an unlearning operator maps it to 𝑀𝑢, scored against the gold-standard retrain 𝑀∗ . The benchmark spans eight task families, twelve methods, and a seven-metric panel covering utility, privacy, backdoor erasure, efficiency, and durability [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. (a) Forget vs retain accuracy on BloodMNIST (easiest request), gentle cluster [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Segmentation case study across three datasets, showing forget (red) and retained [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Case studies on four further task families, each showing two forget (red) and one [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Label heterogeneity of the 20-client partitions across MedMNIST (rows) at Dirichlet 𝛼 ∈ {0.1, 0.5, 1.0} (columns). Each bubble is a (client, class) cell sized by sample count. B.4 Scale and Durability Two controls separate a real forgetting effect from an artifact: enl…
Figure 6
Figure 6. Figure 6: Average rank by retain accuracy across the four MedMNIST client-level datasets [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]
Figure 7
Figure 7. Figure 7: Class-level forgetting on BloodMNIST (forget class [PITH_FULL_IMAGE:figures/full_fig_p024_7.png]
Figure 8
Figure 8. Figure 8: Client-count scaling on BloodMNIST. Top row: naive scaling (forget, retain). Bottom [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Robustness on BloodMNIST. Left: backdoor attack-success-rate after unlearning [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Class-level forgetting on VQA-RAD, t-SNE of the VQANet fused vision-language [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]
Figure 11
Figure 11. Figure 11: Qualitative segmentation case study on the forget client (rows: Kvasir-SEG, [PITH_FULL_IMAGE:figures/full_fig_p029_11.png]
Figure 12
Figure 12. Figure 12: Low-dose-CT denoising case study (AAPM-Mayo-2016). Columns: quarter-dose [PITH_FULL_IMAGE:figures/full_fig_p029_12.png]
Figure 13
Figure 13. Figure 13: Cross-modality synthesis case study (IXI T1 [PITH_FULL_IMAGE:figures/full_fig_p030_13.png]
Figure 14
Figure 14. Figure 14: Registration case study: each method’s warp of the moving image toward the [PITH_FULL_IMAGE:figures/full_fig_p030_14.png]
Figure 15
Figure 15. Figure 15: Lesion-localization case study: the forget client’s predicted lesion centroid per [PITH_FULL_IMAGE:figures/full_fig_p031_15.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 34 canonical work pages

  1. [1]

    InterpretableCNN-multilevelattentiontransformerforrapidrecognitionofpneumonia from chest X-ray images.IEEE Journal of Biomedical and Health Informatics, 28(2):753–764, 2023

    Shengchao Chen, Sufen Ren, Guanjun Wang, Mengxing Huang, and Chenyang Xue. InterpretableCNN-multilevelattentiontransformerforrapidrecognitionofpneumonia from chest X-ray images.IEEE Journal of Biomedical and Health Informatics, 28(2):753–764, 2023

    work page 2023

  2. [2]

    Communication-efficient learning of deep networks from decentralized data

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. Communication-efficient learning of deep networks from decentralized data. InArtificial intelligence and statistics, pages 1273–1282. PMLR, 2017

    2017

  3. [3]

    FLamby: Datasets and benchmarks for cross-silo federated learning in realistic healthcare settings.Advances in Neural Information Processing Systems, 35:5315–5334, 2022

    Jean Ogier du Terrail, Samy-Safwan Ayed, Edwige Cyffers, Felix Grimberg, Chaoyang He, Regis Loeb, Paul Mangold, Tanguy Marchand, Othmane Marfoq, Erum Mushtaq, et al. FLamby: Datasets and benchmarks for cross-silo federated learning in realistic healthcare settings.Advances in Neural Information Processing Systems, 35:5315–5334, 2022

    work page 2022

  4. [4]

    FedEraser: Enabling efficient client-level data removal from federated learning models

    Gaoyang Liu, Xiaoqiang Ma, Yang Yang, Chen Wang, and Jiangchuan Liu. FedEraser: Enabling efficient client-level data removal from federated learning models. In2021 IEEE/ACM 29th international symposium on quality of service (IWQOS), pages 1–10. IEEE, 2021

    work page

  5. [5]

    Federated unlearning: How to efficiently erase a client in fl?arXiv preprint arXiv:2207.05521, 2022

    Anisa Halimi, Swanand Kadhe, Ambrish Rawat, and Nathalie Baracaldo. Federated unlearning: How to efficiently erase a client in fl?arXiv preprint arXiv:2207.05521, 2022

    arXiv 2022

  6. [6]

    Federated unlearning: A survey on methods, design guidelines, and evaluation metrics.IEEE transactions on neural networks and learning systems, pages 1–21, 2024

    NicolòRomandini,AlessioMora,CarloMazzocca,RebeccaMontanari,PaoloBellavista, et al. Federated unlearning: A survey on methods, design guidelines, and evaluation metrics.IEEE transactions on neural networks and learning systems, pages 1–21, 2024

    work page 2024

  7. [7]

    Deep unlearn: Benchmarking machine unlearning for image classification.arXiv preprint arXiv:2410.01276, 2024

    Xavier F Cadet, Anastasia Borovykh, Mohammad Malekzadeh, Sara Ahmadi-Abhari, and Hamed Haddadi. Deep unlearn: Benchmarking machine unlearning for image classification.arXiv preprint arXiv:2410.01276, 2024

    arXiv 2024

  8. [8]

    Enable the right to be forgotten with federated client unlearning in medical imaging

    Zhipeng Deng, Luyang Luo, and Hao Chen. Enable the right to be forgotten with federated client unlearning in medical imaging. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 240–250. Springer, 2024

    work page 2024

Show all 51 references
  1. [9]

    Maverick:Collaboration-freefederatedunlearning formedicalprivacy

    WinKentOngandCheeSengChan. Maverick:Collaboration-freefederatedunlearning formedicalprivacy. InInternationalConferenceonMedicalImageComputingandComputer- Assisted Intervention, pages 358–368. Springer, 2025

    2025

  2. [10]

    Restyled, tuning, and alignment: Taming VLMs for federated non-IID medical image analysis

    Shengchao Chen and Ting Shu. Restyled, tuning, and alignment: Taming VLMs for federated non-IID medical image analysis. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 530–540. Springer, 2025

    2025

  3. [11]

    Taming vision-language models for federated founda- tion models on heterogeneous medical imaging modalities

    Lulu Feng and Shengchao Chen. Taming vision-language models for federated founda- tion models on heterogeneous medical imaging modalities. InProceedings of the 2025 International Conference on Multimedia Retrieval, pages 303–311, 2025

    2025

  4. [12]

    Visual and textual spaces both matter: Taming CLIP fornon-IIDfederatedmedicalimageclassification.Knowledge-BasedSystems,338:115524, 2026

    Lulu Feng and Shengchao Chen. Visual and textual spaces both matter: Taming CLIP fornon-IIDfederatedmedicalimageclassification.Knowledge-BasedSystems,338:115524, 2026

    2026

  5. [13]

    Prompt federated learning for weather forecasting: Toward foundation models on meteorological data.arXiv preprint arXiv:2301.09152, 2023

    Shengchao Chen, Guodong Long, Tao Shen, and Jing Jiang. Prompt federated learning for weather forecasting: Toward foundation models on meteorological data.arXiv preprint arXiv:2301.09152, 2023

    2023 arXiv

  6. [14]

    Fed- erated prompt learning for weather foundation models on devices.arXiv preprint arXiv:2305.14244, 2023

    Shengchao Chen, Guodong Long, Tao Shen, Jing Jiang, and Chengqi Zhang. Fed- erated prompt learning for weather foundation models on devices.arXiv preprint arXiv:2305.14244, 2023. 12

    2023 arXiv

  7. [15]

    Federatedfoundation models on heterogeneous time series

    ShengchaoChen,GuodongLong,JingJiang,andChengqiZhang. Federatedfoundation models on heterogeneous time series. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 15839–15847, 2025

    2025

  8. [16]

    FeDaL: Federated dataset learning for general time series foundation models

    Shengchao Chen, Guodong Long, Michael Blumenstein, and Jing Jiang. FeDaL: Federated dataset learning for general time series foundation models. InInternational Conference on Learning Representations, volume 2026, pages 98850–98885, 2026

    2026

  9. [17]

    Federated unlearning via class- discriminative pruning

    Junxiao Wang, Song Guo, Xin Xie, and Heng Qi. Federated unlearning via class- discriminative pruning. InProceedings of the ACM web conference 2022, pages 622–632, 2022

    2022

  10. [18]

    FedQUIT: On-device federated unlearning via a quasi-competent virtual teacher.arXiv preprint arXiv:2408.07587, 2024

    Alessio Mora, Lorenzo Valerio, Paolo Bellavista, and Andrea Passarella. FedQUIT: On-device federated unlearning via a quasi-competent virtual teacher.arXiv preprint arXiv:2408.07587, 2024

    2024 arXiv

  11. [19]

    Federatedunlearningusingdiffusivenoiseinjection.InformationFusion,page 103796, 2025

    Muhammad Mustafa Ali Usmani, Muhammad Atif Tahir, Humna Faisal, and Muham- madRafi. Federatedunlearningusingdiffusivenoiseinjection.InformationFusion,page 103796, 2025

    2025

  12. [20]

    Unlearning through knowledge overwriting: Reversible federated unlearning via selective sparse adapter

    Zhengyi Zhong, Weidong Bao, Ji Wang, Shuai Zhang, Jingxuan Zhou, Lingjuan Lyu, and Wei Yang Bryan Lim. Unlearning through knowledge overwriting: Reversible federated unlearning via selective sparse adapter. InProceedings of the Computer Vision and Pattern Recognition Conferenc...

    2025

  13. [21]

    FedCARE:Federatedunlearningwithconflict-awareprojectionandrelearning-resistant recovery.arXiv preprint arXiv:2601.22589, 2026

    YueLi,MingminChu,XileiYang,DaXiao,ZiqiXu,WeiShao,QipengSong,andHuiLi. FedCARE:Federatedunlearningwithconflict-awareprojectionandrelearning-resistant recovery.arXiv preprint arXiv:2601.22589, 2026

    2026

  14. [22]

    Membership inference attacks against machine learning models

    Reza Shokri, Marco Stronati, Congzheng Song, and Vitaly Shmatikov. Membership inference attacks against machine learning models. In2017 IEEE symposium on security and privacy (SP), pages 3–18. IEEE, 2017

    2017

  15. [23]

    Comprehensive privacy analysis of deep learning: Passive and active white-box inference attacks against centralized and federated learning

    Milad Nasr, Reza Shokri, and Amir Houmansadr. 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), pages 739–753. IEEE, 2019

    2019

  16. [24]

    Membership inference attacks from first principles

    Nicholas Carlini, Steve Chien, Milad Nasr, Shuang Song, Andreas Terzis, and Florian Tramer. Membership inference attacks from first principles. In2022 IEEE symposium on security and privacy (SP), pages 1897–1914. IEEE, 2022

    1914

  17. [25]

    Inexact unlearning needs more careful evaluations to avoid a false sense of privacy

    Jamie Hayes, Ilia Shumailov, Eleni Triantafillou, Amr Khalifa, and Nicolas Papernot. Inexact unlearning needs more careful evaluations to avoid a false sense of privacy. In2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), pages 497–519. IEEE, 2025

    2025

  18. [26]

    MedMNIST v2-a large-scale lightweight benchmark for 2D and 3D biomedical image classification.Scientific data, 10(1):41, 2023

    Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, and Bingbing Ni. MedMNIST v2-a large-scale lightweight benchmark for 2D and 3D biomedical image classification.Scientific data, 10(1):41, 2023

    2023

  19. [27]

    Kvasir: A multi-class image dataset for computer aided gastrointestinal disease detection

    Konstantin Pogorelov, Kristin Ranheim Randel, Carsten Griwodz, Sigrun Losada Eskeland, Thomas de Lange, Dag Johansen, Concetto Spampinato, Duc-Tien Dang- Nguyen, Mathias Lux, Peter Thelin Schmidt, et al. Kvasir: A multi-class image dataset for computer aided gastrointestinal d...

    2017

  20. [28]

    From detection of individual metastases to classification of lymph node statusatthepatientlevel:thecamelyon17challenge.IEEEtransactionsonmedicalimaging, 38(2):550–560, 2018

    PeterBandi,OscarGeessink,QuirineManson,MarcoryVanDijk,MaschenkaBalkenhol, Meyke Hermsen, Babak Ehteshami Bejnordi, Byungjae Lee, Kyunghyun Paeng, Aoxiao Zhong, et al. From detection of individual metastases to classification of lymph node statusatthepatientlevel:thecamelyon17ch...

    2018

  21. [29]

    WILDS: A benchmark of in-the-wild distribution shifts

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. WILDS: A benchmark of in-the-wild distribution shifts. InInternational conference on machine learning, pa...

    2021

  22. [30]

    Kvasir-SEG: A segmented polyp dataset

    Debesh Jha, Pia H Smedsrud, Michael A Riegler, Pål Halvorsen, Thomas De Lange, Dag Johansen, and Håvard D Johansen. Kvasir-SEG: A segmented polyp dataset. In International conference on multimedia modeling, pages 451–462. Springer, 2019

    2019

  23. [31]

    WM-DOVA maps for accurate polyp highlighting in colonoscopy: Validation vs

    Jorge Bernal, F Javier Sánchez, Gloria Fernández-Esparrach, Debora Gil, Cristina Rodríguez, and Fernando Vilariño. WM-DOVA maps for accurate polyp highlighting in colonoscopy: Validation vs. saliency maps from physicians.Computerized medical imaging and graphics, 43:99–111, 2015

    2015

  24. [32]

    Skin lesion analysis toward melanoma detection: A challenge at the 2017 internationalsymposiumonbiomedicalimaging(isbi),hostedbytheinternationalskin imaging collaboration (isic)

    Noel CF Codella, David Gutman, M Emre Celebi, Brian Helba, Michael A Marchetti, Stephen W Dusza, Aadi Kalloo, Konstantinos Liopyris, Nabin Mishra, Harald Kittler, et al. Skin lesion analysis toward melanoma detection: A challenge at the 2017 internationalsymposiumonbiomedicali...

    2017

  25. [33]

    McCollough, Adam C

    Cynthia H. McCollough, Adam C. Bartley, Rickey E. Carter, Baiyu Chen, Tammy A. Drees, Phillip Edwards, David R. Holmes III, Alice E. Huang, Farhana Khan, Shuai Leng, Kyle L. McMillan, Gregory J. Michalak, Kristina M. Nunez, Lifeng Yu, and Joel G. Fletcher. Low-dose CT for the ...

    2016

  26. [34]

    IXI dataset — informa- tion extraction from images.https://brain-development.org/ixi-dataset/, 2007

    Biomedical Image Analysis Group, Imperial College London. IXI dataset — informa- tion extraction from images.https://brain-development.org/ixi-dataset/, 2007. T1/T2 brain MRI from three London hospitals, CC BY-SA 3.0

    2007

  27. [35]

    A dataset of clinically generated visual questions and answers about radiology images.Scientific data, 5(1):180251, 2018

    Jason J Lau, Soumya Gayen, Asma Ben Abacha, and Dina Demner-Fushman. A dataset of clinically generated visual questions and answers about radiology images.Scientific data, 5(1):180251, 2018

    2018

  28. [36]

    SLAKE: A semantically-labeled knowledge-enhanced dataset for medical visual question answer- ing

    Bo Liu, Li-Ming Zhan, Li Xu, Lin Ma, Yan Yang, and Xiao-Ming Wu. SLAKE: A semantically-labeled knowledge-enhanced dataset for medical visual question answer- ing. In2021 IEEE 18th international symposium on biomedical imaging (ISBI), pages 1650–1654. IEEE, 2021

    2021

  29. [37]

    PathVQA: 30000+ questions for medical visual question answering.arXiv preprint arXiv:2003.10286, 2020

    Xuehai He, Yichen Zhang, Luntian Mou, Eric Xing, and Pengtao Xie. PathVQA: 30000+ questions for medical visual question answering.arXiv preprint arXiv:2003.10286, 2020

    2003 arXiv

  30. [38]

    Measuring the effects of non-identical data distribution for federated visual classification.arXiv preprint arXiv:1909.06335, 2019

    Tzu-Ming Harry Hsu, Hang Qi, and Matthew Brown. Measuring the effects of non-identical data distribution for federated visual classification.arXiv preprint arXiv:1909.06335, 2019

    1909 arXiv

  31. [39]

    Federated unlearning made practical: Seamless integration via negated pseudo-gradients.IEEE Transactions on Big Data, 2026

    Alessio Mora, Carlo Mazzocca, Rebecca Montanari, and Paolo Bellavista. Federated unlearning made practical: Seamless integration via negated pseudo-gradients.IEEE Transactions on Big Data, 2026

    2026

  32. [40]

    LefengZhang,TianqingZhu,HaibinZhang,PingXiong,andWanleiZhou.FedRecovery: Differentially private machine unlearning for federated learning frameworks.IEEE Transactions on Information Forensics and Security, 18:4732–4746, 2023

    2023

  33. [41]

    BFU:Bayesianfederated unlearning with parameter self-sharing

    WeiqiWang,ZhiyiTian,ChenhanZhang,AnLiu,andShuiYu. BFU:Bayesianfederated unlearning with parameter self-sharing. InProceedings of the 2023 ACM Asia Conference on Computer and Communications Security, pages 567–578, 2023. 14

    2023

  34. [42]

    NoT: Federated unlearning via weight negation

    YasserHKhalil,LeoBrunswic,SoufianeLamghari,XuLi,MahdiBeitollahi,andXiChen. NoT: Federated unlearning via weight negation. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 25759–25769, 2025

    2025

  35. [43]

    Deep residual learning for image recognition: A survey.Applied sciences, 12(18):8972, 2022

    Muhammad Shafiq and Zhaoquan Gu. Deep residual learning for image recognition: A survey.Applied sciences, 12(18):8972, 2022

    2022

  36. [44]

    U-net:Convolutionalnetworksfor biomedical image segmentation

    OlafRonneberger,PhilippFischer,andThomasBrox. U-net:Convolutionalnetworksfor biomedical image segmentation. InInternational Conference on Medical image computing and computer-assisted intervention, pages 234–241. Springer, 2015

    2015

  37. [45]

    DistilBERT, a dis- tilledversionofBERT:smaller,faster,cheaperandlighter.arXivpreprintarXiv:1910.01108, 2019

    Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. DistilBERT, a dis- tilledversionofBERT:smaller,faster,cheaperandlighter.arXivpreprintarXiv:1910.01108, 2019

    1910 arXiv

  38. [46]

    VoxelMorph: a learning framework for deformable medical image registration.arXiv preprint arXiv:1809.05231, 2018

    Guha Balakrishnan, Amy Zhao, Mert R Sabuncu, John Guttag, and Adrian V Dalca. VoxelMorph: a learning framework for deformable medical image registration.arXiv preprint arXiv:1809.05231, 2018

    2018 arXiv

  39. [47]

    Personalizedadapter forlargemeteorologymodelondevices:Towardsweatherfoundationmodels.Advances in Neural Information Processing Systems, 37:84897–84943, 2024

    ShengchaoChen,GuodongLong,JingJiang,andChengqiZhang. Personalizedadapter forlargemeteorologymodelondevices:Towardsweatherfoundationmodels.Advances in Neural Information Processing Systems, 37:84897–84943, 2024

    2024

  40. [48]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  41. [49]

    Statistical comparisons of classifiers over multiple data sets.Journal of Machine learning research, 7(Jan):1–30, 2006

    Janez Demšar. Statistical comparisons of classifiers over multiple data sets.Journal of Machine learning research, 7(Jan):1–30, 2006

    2006

  42. [50]

    Visualizing data using t-sne.Journal of Machine Learning Research, 9:2579–2605, 2008

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of Machine Learning Research, 9:2579–2605, 2008. 15 Appendix A Methods and Hyperparameters...................................................16 A.1 Unlearning Methods.................................

    2008

  43. [2017]

    NIH-AAPM-Mayo Clinic Low Dose CT Grand Challenge

Pith tools

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