pith. sign in

arxiv: 2605.22589 · v1 · pith:63LJ7UY5new · submitted 2026-05-21 · 💻 cs.NI

SCALE: Sensitivity-Aware Federated Unlearning with Information Freshness Optimization for Mobile Edge Computing

Zihao Ding , Beining Wu , Jun Huang This is my paper

Pith reviewed 2026-05-22 01:46 UTC · model grok-4.3

classification 💻 cs.NI
keywords federated unlearningmobile edge computinglayer sensitivity analysisadaptive sparsificationinformation freshnessprivacy regulationsconvergence analysis
0
0 comments X

The pith

SCALE achieves more precise federated unlearning by analyzing historical contributions to select sensitive layers and applying freshness-aware sparsification.

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

The paper sets out to show that current federated unlearning methods suffer from imprecise forgetting and overlook time-based information, which hurts performance in mobile edge computing. SCALE fixes this with a two-stage process: it first uses historical client contributions to pinpoint the model layers most affected by the data to be removed, then applies adaptive sparsification to weight subgroups while accounting for how recent the information is. Theoretical work establishes that the method converges and runs faster than baselines. Experiments on both simulations and real testbeds report stronger forgetting results than prior techniques. A reader would care because this helps federated systems meet strict privacy rules without losing overall model quality.

Core claim

The authors establish that a dual-level unlearning process, built on historical contribution-based layer sensitivity analysis to locate affected layers followed by adaptive sparsification at the weight sub-group level with information freshness optimization, delivers higher-precision client data removal, maintains convergence, and yields measurably better forgetting than existing approaches in mobile edge computing.

What carries the argument

Dual-level unlearning mechanism that first runs historical contribution-based layer sensitivity analysis to identify target-influenced layers, then performs adaptive sparsification at the weight sub-group level to trade off information freshness against forgetting effectiveness.

Load-bearing premise

The assumption that a historical contribution-based layer sensitivity analysis can reliably identify the layers most influenced by target clients and that adaptive sparsification at the weight sub-group level can balance information freshness with effective forgetting without introducing new biases or performance drops.

What would settle it

After running the unlearning procedure, measure whether a held-out test set from the target client still produces high model accuracy or allows data reconstruction; if accuracy does not drop relative to a non-unlearned model, the forgetting claim fails.

Figures

Figures reproduced from arXiv: 2605.22589 by Beining Wu, Jun Huang, Zihao Ding.

Figure 1
Figure 1. Figure 1: Client unlearning in the MEC system. global dataset is D = SN n=1 Dn with total size M = |D| = PN n=1 Mn. We assume that the global model consists of L layers, where layer l ∈ {1, 2, . . . , L} contains parameters Wl ∈ R dl with dimension dl . The complete global model parameters are Θ = {W1, W2, . . . , WL} ∈ R d , where d = PL l=1 dl . We define the local loss function for client n as: Fn(Θ) = 1 Mn X Mn … view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the SCALE. B. Problem Formulation Clients can request the removal of their data from the trained global model, which creates challenges due to distributed data ownership, historical contributions over training rounds, and the computational costs associated with full retraining. Assume an unlearning request is specified by U = (Cu, Du, τ ), where Cu ⊆ {1, 2, . . . , N} is the set of requestin… view at source ↗
Figure 3
Figure 3. Figure 3: Testbed setup for real-time MEC FU validation. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: PPO training convergence for different unlearning scenarios. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Layer sensitivity analysis over different model architectures and unlearning scenarios. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sum AoI comparison for different unlearning scenarios across model architectures. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Impact of data heterogeneity on testbed unlearning performance in three scenarios. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
read the original abstract

Federated Unlearning (FU) is emerging as a powerful tool that enables the selective removal of client data to effectively address data contamination and meet strict privacy regulations in mobile edge computing (MEC) systems. Although FU has recently drawn attention in the AI community, existing approaches suffer from low unlearning precision and lack temporal information reflection, which results in suboptimal forgetting performance. To address these issues, we propose SCALE, a dual-level unlearning framework combining historical contribution analysis with information freshness-aware adaptive sparsification. Our framework first employs a historical contribution-based layer sensitivity analysis to identify layers most influenced by target clients, then performs fine-grained unlearning through adaptive sparsification at the weight sub-group level to balance information freshness with forgetting effectiveness. Through theoretical analysis, the proposed framework demonstrates the convergence properties and acceleration advantages. Our experiments and testbed results demonstrate superior unlearning effectiveness compared to state-of-the-art baselines, with significantly improved forgetting performance.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The manuscript proposes SCALE, a dual-level federated unlearning framework for mobile edge computing. It first applies historical contribution-based layer sensitivity analysis to identify layers most influenced by target clients, then performs adaptive sparsification at the weight sub-group level that incorporates information freshness to balance forgetting effectiveness with model utility. The paper claims to establish convergence properties and acceleration advantages via theoretical analysis, and reports superior unlearning effectiveness and forgetting performance over state-of-the-art baselines in experiments and testbed evaluations.

Significance. If the convergence analysis is rigorous and the experimental gains hold under proper controls, the work could meaningfully advance federated unlearning by adding temporal freshness awareness and fine-grained sparsification, potentially yielding more precise forgetting in MEC privacy scenarios without large accuracy penalties. The combination of contribution-driven sensitivity and freshness-aware adaptation is a plausible incremental improvement over existing FU methods.

major comments (2)
  1. [Abstract / Theoretical Analysis] Abstract and theoretical analysis section: The central claim that the dual-level construction (historical contribution sensitivity plus freshness-aware sparsification) yields both convergence and acceleration is load-bearing, yet the abstract supplies no equations, proof sketch, or assumption list. For non-convex deep models it is not obvious that the contribution metric ranks layers by true target-client influence or that the sparsity schedule preserves the underlying optimizer's contraction properties; a mismatch would invalidate both the convergence guarantee and the claimed forgetting improvement.
  2. [Experiments / Testbed Results] Experimental section: The abstract asserts superior unlearning effectiveness and significantly improved forgetting performance, but the provided text contains no tables, error bars, statistical tests, or ablation results on the free parameters (layer sensitivity threshold, freshness weighting factor). Without these, it is impossible to assess whether the gains are robust or merely post-hoc tuning artifacts.
minor comments (1)
  1. [Abstract] The abstract introduces 'information freshness' and 'historical contribution' without even a high-level formula or reference to their definitions; adding a brief notational preview would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment point by point below, providing clarifications from the full manuscript and indicating planned revisions to improve clarity and completeness.

read point-by-point responses

  1. Referee: [Abstract / Theoretical Analysis] Abstract and theoretical analysis section: The central claim that the dual-level construction (historical contribution sensitivity plus freshness-aware sparsification) yields both convergence and acceleration is load-bearing, yet the abstract supplies no equations, proof sketch, or assumption list. For non-convex deep models it is not obvious that the contribution metric ranks layers by true target-client influence or that the sparsity schedule preserves the underlying optimizer's contraction properties; a mismatch would invalidate both the convergence guarantee and the claimed forgetting improvement.

    Authors: We appreciate the referee's emphasis on the importance of clearly conveying the theoretical foundations. The full manuscript (Section 4) provides a rigorous convergence analysis under standard assumptions including L-smoothness of the loss, bounded gradient variance, and unbiased stochastic gradients. The historical contribution metric is formally defined as the normalized cumulative gradient norm from target clients, and we prove it ranks layers by influence through a sensitivity bound. The adaptive sparsification schedule is shown to preserve the underlying optimizer's contraction mapping by bounding the additional error term introduced at each sub-group level, leading to an accelerated convergence rate compared to uniform unlearning baselines. We agree, however, that the abstract would benefit from a concise statement of key assumptions and a high-level proof outline. We will revise the abstract to incorporate this information. revision: yes

  2. Referee: [Experiments / Testbed Results] Experimental section: The abstract asserts superior unlearning effectiveness and significantly improved forgetting performance, but the provided text contains no tables, error bars, statistical tests, or ablation results on the free parameters (layer sensitivity threshold, freshness weighting factor). Without these, it is impossible to assess whether the gains are robust or merely post-hoc tuning artifacts.

    Authors: We acknowledge the referee's valid concern about the presentation of experimental evidence. The full manuscript includes comparative tables (Tables 1–3) reporting unlearning effectiveness and forgetting performance metrics averaged over 5 independent runs, with error bars shown in the corresponding figures. Ablation studies varying the layer sensitivity threshold and freshness weighting factor appear in Section 5.3, demonstrating that performance remains stable across a range of values. To strengthen the manuscript, we will add explicit statistical significance tests (paired t-tests with p-values) to the tables and expand the ablation discussion in the main text to further address robustness concerns. revision: yes

Circularity Check

0 steps flagged

No significant circularity; theoretical claims presented as independent of fitted inputs

full rationale

The abstract and provided context describe a dual-level framework using historical contribution-based layer sensitivity analysis followed by adaptive sparsification, with convergence and acceleration shown via theoretical analysis. No equations, self-citations, or fitted parameters are visible that reduce any prediction or uniqueness claim to the inputs by construction. The derivation chain remains self-contained against external benchmarks, as the central claims rest on modeling steps that are not shown to be tautological or statistically forced within the given text.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Abstract-only view limits visibility; the method implicitly relies on standard federated convergence assumptions and likely introduces tunable thresholds for sensitivity and sparsification that are not enumerated.

free parameters (2)
  • layer sensitivity threshold
    Used to select layers for unlearning; value would be fitted or chosen to balance forgetting and accuracy.
  • freshness weighting factor
    Controls how information age affects sparsification; appears as an adaptive parameter.
axioms (1)
  • standard math Federated learning updates converge under standard assumptions on client participation and learning rates
    Invoked to support the claimed convergence properties of the unlearning process.

pith-pipeline@v0.9.0 · 5691 in / 1353 out tokens · 58591 ms · 2026-05-22T01:46:23.322805+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We develop a theoretical framework for SCALE by demonstrating the convergence properties and acceleration advantages of SCALE. The analytical results show that the designed framework achieves an O(√(L/|Ls|)) convergence advantage over the uniform parameter modification framework

  • IndisputableMonolith/Foundation/ArrowOfTime.lean arrow_from_z unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    AoI-driven adaptive sparsification... reward R[t] = wf Rf[t] + wc Rc[t] with forgetting and freshness terms

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 5 internal anchors

  1. [1]

    Enhancing Vehicular Platooning With Wireless Federated Learning: A Resource-Aware Con- trol Framework,

    B. Wu, J. Huang, Q. Duan, L. Dong, and Z. Cai, “Enhancing Vehicular Platooning With Wireless Federated Learning: A Resource-Aware Con- trol Framework,”IEEE Transactions on Networking, pp. 1–16, 2025

    work page 2025

  2. [2]

    Forget Unlearning: Towards True Data- Deletion in Machine Learning,

    R. Chourasia and N. Shah, “Forget Unlearning: Towards True Data- Deletion in Machine Learning,” inInternational Conference on Machine Learning. PMLR, 2023, pp. 6028–6073

    work page 2023

  3. [3]

    An Efficient and Multi-Dimensional Privacy-Preserving Platoon Communi- cation Scheme in Vehicular Networks,

    N. Xu, Z. Liu, X. Han, Q. Guan, H. Sun, X. Huang, and J. Ma, “An Efficient and Multi-Dimensional Privacy-Preserving Platoon Communi- cation Scheme in Vehicular Networks,”IEEE Transactions on Intelligent Transportation Systems, vol. 26, no. 5, pp. 6831–6847, 2025

    work page 2025

  4. [4]

    RELIEF: Turning Missing Modalities into Training Acceleration for Federated Learning on Heterogeneous IoT Edge

    B. Wu, Z. Ding, and J. Huang, “RELIEF: Turning Missing Modalities into Training Acceleration for Federated Learning on Heterogeneous IoT Edge,” arXiv preprint arXiv:2604.04243, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  5. [5]

    Transformer-Based Dynamic Resource Allocation for Multi-Carrier NOMA Systems,

    L. Dong, J. Huang, and R. W. Heath, “Transformer-Based Dynamic Resource Allocation for Multi-Carrier NOMA Systems,”IEEE Transac- tions on Cognitive Communications and Networking, vol. 12, pp. 4926– 4941, 2026

    work page 2026

  6. [6]

    Federated Unlearning: A Survey on Methods, Design Guidelines, and Evaluation Metrics,

    N. Romandini, A. Mora, C. Mazzocca, R. Montanari, and P. Bellavista, “Federated Unlearning: A Survey on Methods, Design Guidelines, and Evaluation Metrics,”IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 7, pp. 11 697–11 717, 2025

    work page 2025

  7. [7]

    FedTD3: An Accelerated Learning Approach for UA V Trajectory Planning,

    B. Wu, J. Huang, and Q. Duan, “FedTD3: An Accelerated Learning Approach for UA V Trajectory Planning,” inInternational Conference on Wireless Artificial Intelligent Computing Systems and Applications (WASA). Springer, 2025, pp. 13–24

    work page 2025

  8. [8]

    A Dual- Level Game-Theoretic Approach for Collaborative Learning in UA V- Assisted Heterogeneous Vehicle Networks,

    Z. Ding, J. Huang, Q. Duan, C. Zhang, Y . Zhao, and S. Gu, “A Dual- Level Game-Theoretic Approach for Collaborative Learning in UA V- Assisted Heterogeneous Vehicle Networks,” in2025 IEEE International Performance, Computing, and Communications Conference (IPCCC). IEEE, 2025, pp. 1–8

    work page 2025

  9. [9]

    BFU: Bayesian Federated Unlearning With Parameter Self-Sharing,

    W. Wang, Z. Tian, C. Zhang, A. Liu, and S. Yu, “BFU: Bayesian Federated Unlearning With Parameter Self-Sharing,” inProc. ACM Asia Conf. on Computer and Communications Security (AsiaCCS), 2023, pp. 567–578

    work page 2023

  10. [10]

    Federaser: Enabling Efficient Client-Level Data Removal From Federated Learning Models,

    G. Liu, X. Ma, Y . Yang, C. Wang, and J. Liu, “Federaser: Enabling Efficient Client-Level Data Removal From Federated Learning Models,” inProc. IEEE/ACM 29th Int. Symp. on Quality of Service (IWQoS). IEEE, 2021, pp. 1–10

    work page 2021

  11. [11]

    The Right to Be Forgotten in Federated Learning: An Efficient Realization With Rapid Retraining,

    Y . Liu, L. Xu, X. Yuan, C. Wang, and B. Li, “The Right to Be Forgotten in Federated Learning: An Efficient Realization With Rapid Retraining,” inProc. IEEE Conf. on Computer Communications (INFOCOM). IEEE, 2022, pp. 1749–1758

    work page 2022

  12. [12]

    Real-Time Intelligent Healthcare Enabled by Federated Digital Twins With AoI Optimization,

    B. Wu, J. Huang, and Q. Duan, “Real-Time Intelligent Healthcare Enabled by Federated Digital Twins With AoI Optimization,”IEEE Network, vol. 40, no. 2, pp. 184–191, 2026

    work page 2026

  13. [13]

    Reinforcement Learning- Based Energy-Aware Coverage Path Planning for Precision Agriculture,

    B. Wu, Z. Ding, L. Ostigaard, and J. Huang, “Reinforcement Learning- Based Energy-Aware Coverage Path Planning for Precision Agriculture,” in2025 ACM Research on Adaptive and Convergent Systems (RACS). ACM, 2025, pp. 1–8

    work page 2025

  14. [14]

    AoI-aware Resource Management for Smart Health via Deep Reinforcement Learning,

    B. Wu, Z. Cai, W. Wu, and X. Yin, “AoI-aware Resource Management for Smart Health via Deep Reinforcement Learning,”IEEE Access, 2023

    work page 2023

  15. [15]

    AoI-Aware Federated Unlearning for Streaming Data With Online Client Selection and Pricing,

    Y . Cui, N. Ding, and M. H. Cheung, “AoI-Aware Federated Unlearning for Streaming Data With Online Client Selection and Pricing,” in IEEE INFOCOM 2025–IEEE Conference on Computer Communica- tions. IEEE, 2025, pp. 1–10

    work page 2025

  16. [16]

    A Stochastic Geometry- Based Analysis of SWIPT-Assisted Underlaid Device-to-Device Energy Harvesting,

    C.-C. Xing, Z. Ding, and J. Huang, “A Stochastic Geometry- Based Analysis of SWIPT-Assisted Underlaid Device-to-Device Energy Harvesting,”SIGAPP Appl. Comput. Rev., vol. 25, no. 4, p. 18–34, Jan

    work page

  17. [17]

    Available: https://doi.org/10.1145/3787594.3787596

    [Online]. Available: https://doi.org/10.1145/3787594.3787596

    work page doi:10.1145/3787594.3787596

  18. [18]

    Hybrid Edge Intelligence for Real- Time Intrusion Detection in Advanced Traffic Management Systems,

    R. R. Depa, Y . Zhang, and D. Xu, “Hybrid Edge Intelligence for Real- Time Intrusion Detection in Advanced Traffic Management Systems,” in2025 IEEE Security and Privacy Workshops (SPW). IEEE, 2025, pp. 352–354

    work page 2025

  19. [19]

    Learning to Defend: A Multi-Agent Reinforcement Learning Framework for Stackelberg Security Game in Mobile Edge Computing,

    Z. Ding, J. Huang, and J. Qi, “Learning to Defend: A Multi-Agent Reinforcement Learning Framework for Stackelberg Security Game in Mobile Edge Computing,” in2026 International Conference on Computing, Networking and Communications (ICNC), 2026, pp. 769– 774

    work page 2026

  20. [20]

    Federated Unlearning With Fast Recovery,

    C. Zhou, C. Pan, M. Li, and P. Wang, “Federated Unlearning With Fast Recovery,”IEEE Transactions on Mobile Computing, pp. 1–18, 2025

    work page 2025

  21. [21]

    Exploring Federated Unlearning: Review, Comparison, and Insights,

    Y . Zhao, J. Yang, Y . Tao, L. Wang, X. Li, D. Niyato, and H. Vin- cent Poor, “Exploring Federated Unlearning: Review, Comparison, and Insights,”IEEE Network, pp. 1–1, 2025

    work page 2025

  22. [22]

    Robust Federated Unlearning,

    X. Sheng, W. Bao, and L. Ge, “Robust Federated Unlearning,” in Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (CIKM), 2024, pp. 2034–2044

    work page 2024

  23. [23]

    Certified Unlearning for Federated Recommendation,

    T. T. Huynh, T. B. Nguyen, T. T. Nguyen, P. L. Nguyen, H. Yin, Q. V . H. Nguyen, and T. T. Nguyen, “Certified Unlearning for Federated Recommendation,”ACM Transactions on Information Systems, vol. 43, no. 2, pp. 1–29, 2025

    work page 2025

  24. [24]

    FU-PA: Federated Unlearning via Parameters Adjustment,

    W. Zeng, S. Chen, X. Li, and S. Chen, “FU-PA: Federated Unlearning via Parameters Adjustment,”IEEE Transactions on Emerging Topics in Computational Intelligence, pp. 1–12, 2025

    work page 2025

  25. [25]

    Speed Up Federated Unlearning With Temporary Local Models,

    M. Ameen, P. Wang, W. Su, X. Wei, and Q. Zhang, “Speed Up Federated Unlearning With Temporary Local Models,”IEEE Transactions on Sustainable Computing, pp. 1–16, 2025

    work page 2025

  26. [26]

    Machine Unlearning,

    L. Bourtoule, V . Chandrasekaran, C. A. Choquette-Choo, H. Jia, A. Travers, B. Zhang, D. Lie, and N. Papernot, “Machine Unlearning,” in2021 IEEE Symposium on Security and Privacy (SP). IEEE, 2021, pp. 141–159

    work page 2021

  27. [27]

    Federated Unlearning via Class- Discriminative Pruning,

    J. Wang, S. Guo, X. Xie, and H. Qi, “Federated Unlearning via Class- Discriminative Pruning,” inProceedings of the ACM Web Conference 2022, 2022, pp. 622–632

    work page 2022

  28. [28]

    Proof of Unlearning: Definitions and Instantiation,

    J. Weng, S. Yao, Y . Du, J. Huang, J. Weng, and C. Wang, “Proof of Unlearning: Definitions and Instantiation,”IEEE Transactions on Information Forensics and Security, vol. 19, pp. 3309–3323, 2024

    work page 2024

  29. [29]

    Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning,

    E. Chien, H. Wang, Z. Chen, and P. Li, “Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning,” Advances in Neural Information Processing Systems (NeurIPS), vol. 37, pp. 79 666–79 703, 2024

    work page 2024

  30. [30]

    A Fast UA V Tra- jectory Planning Framework in RIS-Assisted Communication Systems With Accelerated Learning via Multithreading and Federating,

    J. Huang, B. Wu, Q. Duan, L. Dong, and S. Yu, “A Fast UA V Tra- jectory Planning Framework in RIS-Assisted Communication Systems With Accelerated Learning via Multithreading and Federating,”IEEE Transactions on Mobile Computing, pp. 1–16, 2025

    work page 2025

  31. [31]

    Unified Gradient-Based Machine Unlearning With Remain Geometry Enhancement,

    Z. Huang, X. Cheng, J. Zheng, H. Wang, Z. He, T. Li, and X. Huang, “Unified Gradient-Based Machine Unlearning With Remain Geometry Enhancement,”Advances in Neural Information Processing Systems (NeurIPS), vol. 37, pp. 26 377–26 414, 2024

    work page 2024

  32. [32]

    GDR-GMA: Machine Unlearning via Direction-Rectified and Magnitude-Adjusted Gradients,

    S. Lin, X. Zhang, W. Susilo, X. Chen, and J. Liu, “GDR-GMA: Machine Unlearning via Direction-Rectified and Magnitude-Adjusted Gradients,” inProceedings of the 32nd ACM International Conference on Multimedia (ACM MM), 2024, pp. 9087–9095

    work page 2024

  33. [33]

    Federated Unlearning With Gradient Descent and Conflict Mitigation,

    Z. Pan, Z. Wang, C. Li, K. Zheng, B. Wang, X. Tang, and J. Zhao, “Federated Unlearning With Gradient Descent and Conflict Mitigation,” inProceedings of the AAAI Conference on Artificial Intelligence (AAAI), vol. 39, no. 19, 2025, pp. 19 804–19 812

    work page 2025

  34. [34]

    PRISM: Exposing and Resolving Spurious Isolation in Federated Multimodal Continual Learning

    B. Wu, Z. Ding, and J. Huang, “PRISM: Exposing and Resolving Spurious Isolation in Federated Multimodal Continual Learning,” arXiv preprint arXiv:2605.01061, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  35. [35]

    Balancing Forget Quality and Model Utility: A Reverse KL-Divergence Knowledge Distillation Approach for Better Unlearning in LLMs,

    B. Wang, Y . Zi, Y . Sun, Y . Zhao, and B. Qin, “Balancing Forget Quality and Model Utility: A Reverse KL-Divergence Knowledge Distillation Approach for Better Unlearning in LLMs,” inProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers...

    work page 2025

  36. [36]

    Layer Attack Unlearning: Fast and Accurate Machine Unlearning via Layer Level Attack and Knowledge Distillation,

    H. Kim, S. Lee, and S. S. Woo, “Layer Attack Unlearning: Fast and Accurate Machine Unlearning via Layer Level Attack and Knowledge Distillation,” inProceedings of the AAAI Conference on Artificial Intelligence (AAAI), vol. 38, no. 19, 2024, pp. 21 241–21 248

    work page 2024

  37. [37]

    Update Selective Pa- rameters: Federated Machine Unlearning Based on Model Explanation,

    H. Xu, T. Zhu, L. Zhang, W. Zhou, and P. S. Yu, “Update Selective Pa- rameters: Federated Machine Unlearning Based on Model Explanation,” IEEE Transactions on Big Data, vol. 11, no. 2, pp. 524–539, 2025

    work page 2025

  38. [38]

    Sky of Unlearning (SoUL): Rewiring Federated Machine Unlearning via Selective Pruning,

    M. M. U. Zaman, X. Sun, and J. Yao, “Sky of Unlearning (SoUL): Rewiring Federated Machine Unlearning via Selective Pruning,”arXiv preprint arXiv:2504.01705 (arXiv), 2025

    work page arXiv 2025

  39. [39]

    Towards Safe Machine Unlearning: A Paradigm That Mitigates Performance Degradation,

    S. Ye, J. Lu, and G. Zhang, “Towards Safe Machine Unlearning: A Paradigm That Mitigates Performance Degradation,” inProceedings of the ACM on Web Conference 2025, 2025, pp. 4635–4652

    work page 2025

  40. [40]

    Machine Unlearning Through Fine-Grained Model Parameters Perturbation,

    Z. Zuo, Z. Tang, K. Li, and A. Datta, “Machine Unlearning Through Fine-Grained Model Parameters Perturbation,”IEEE Transactions on Knowledge and Data Engineering, vol. 37, no. 4, pp. 1975–1988, 2025

    work page 1975

  41. [41]

    A Review of Continual Learning in Edge AI,

    B. Wu, Z. Ding, and J. Huang, “A Review of Continual Learning in Edge AI,”IEEE Transactions on Network Science and Engineering, vol. 13, pp. 6571–6588, 2026

    work page 2026

  42. [42]

    Dynamic Graph Unlearning: A General and Efficient Post-Processing Method via Gradient Transformation,

    H. Zhang, B. Wu, X. Yang, X. Yuan, X. Liu, and X. Yi, “Dynamic Graph Unlearning: A General and Efficient Post-Processing Method via Gradient Transformation,” inProceedings of the ACM on Web Conference 2025, 2025, pp. 931–944

    work page 2025

  43. [43]

    Lifecycle-Aware Federated Continual Learning in Mobile Autonomous Systems

    B. Wu and J. Huang, “Lifecycle-Aware Federated Continual Learning in Mobile Autonomous Systems,” arXiv preprint arXiv:2604.20745, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  44. [44]

    EASE: Federated Multimodal Unlearning via Entanglement-Aware Anchor Closure

    Z. Ding, B. Wu, and J. Huang, “EASE: Federated Multimodal Un- learning via Entanglement-Aware Anchor Closure,” arXiv preprint arXiv:2605.00733, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  45. [45]

    Incentivized Federated Learning and Unlearning,

    N. Ding, Z. Sun, E. Wei, and R. Berry, “Incentivized Federated Learning and Unlearning,”IEEE Transactions on Mobile Computing, pp. 1–17, 2025

    work page 2025

  46. [46]

    Application-Aware Twin-in-the-Loop Planning for Federated Split Learning over Wireless Edge Networks

    Z. Ding, B. Wu, J. Huang, and S. Mao, “Application-Aware Twin-in- the-Loop Planning for Federated Split Learning over Wireless Edge Networks,” arXiv preprint arXiv:2604.26105, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  47. [47]

    “X of Information

    B. Wu, J. Huang, and S. Yu, ““X of Information” Continuum: A Survey on AI-Driven Multi-Dimensional Metrics for Next-Generation Net- worked Systems,”IEEE Communications Surveys & Tutorials, vol. 28, pp. 5307–5344, 2026

    work page 2026

  48. [48]

    Securing Smart Agriculture with Communication-Efficient Federated Unlearning,

    U. Pudasaini, Z. Ding, and J. Huang, “Securing Smart Agriculture with Communication-Efficient Federated Unlearning,” in2026 IEEE International Conference on High Performance Switching and Routing (HPSR). IEEE, 2026, pp. 1–8

    work page 2026

  49. [49]

    Pearson’s correlation coefficient,

    P. Sedgwick, “Pearson’s correlation coefficient,”Bmj, vol. 345, 2012

    work page 2012

  50. [50]

    R ´enyi Divergence and Kullback-Leibler Divergence,

    T. van Erven and P. Harremos, “R ´enyi Divergence and Kullback-Leibler Divergence,”IEEE Transactions on Information Theory, vol. 60, no. 7, pp. 3797–3820, 2014

    work page 2014

  51. [51]

    Learn to Forget: Machine Unlearning via Neuron Masking,

    Z. Ma, Y . Liu, X. Liu, J. Liu, J. Ma, and K. Ren, “Learn to Forget: Machine Unlearning via Neuron Masking,”IEEE Transactions on De- pendable and Secure Computing, vol. 20, no. 4, pp. 3194–3207, 2023

    work page 2023

  52. [52]

    FedEraser: Enabling Efficient Client-Level Data Removal from Federated Learning Models,

    G. Liu, X. Ma, Y . Yang, C. Wang, and J. Liu, “FedEraser: Enabling Efficient Client-Level Data Removal from Federated Learning Models,” in2021 IEEE/ACM 29th International Symposium on Quality of Service (IWQOS), 2021, pp. 1–10

    work page 2021

  53. [53]

    Federated unlearning: How to effi- ciently erase a client in fl? arXiv preprint arXiv:2207.05521,

    A. Halimi, S. Kadhe, A. Rawat, and N. Baracaldo, “Federated Unlearning: How to Efficiently Erase a Client in FL?” 2023. [Online]. Available: https://arxiv.org/abs/2207.05521

    work page arXiv 2023

  54. [54]

    Unlearning Through Knowledge Overwriting: Reversible Feder- ated Unlearning via Selective Sparse Adapter,

    Z. Zhong, W. Bao, J. Wang, S. Zhang, J. Zhou, L. Lyu, and W. Y . B. Lim, “Unlearning Through Knowledge Overwriting: Reversible Feder- ated Unlearning via Selective Sparse Adapter,” inProceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025, pp. 30 661–30 670

    work page 2025