betaFeedback

WorksComputer Science

Do unlearning algorithms learn the right features?

Abstract

Machine unlearning is commonly evaluated with standard post-deletion metrics that compare an unlearned model to one retrained on the remaining data. Such snapshot evaluations can conceal consequential differences in the features the two models encode. We prove that if an unlearned and a retrained model agree on their outputs but encode different features, a single gradient step on their prediction heads can break this agreement. We further show that deleting an arbitrarily small fraction of the data can cause retraining to select qualitatively different features, so matching retraining can require learning features that the full model never had. We then introduce a measure of feature recovery that separates missing from extraneous features. Across eight unlearning methods, models that perform well on standard metrics often fail to recover the retrained features. This mismatch predicts deviations from retraining under subsequent relearning and downstream adaptation, and directly correcting it systematically reduces these deviations.

Introduction

Training data may need to be removed after a model has been trained, for example, in response to privacy requests, licensing issues, or discovered errors. Deleting these examples from the dataset, however, does not remove their influence from the trained model. Retraining from scratch on the remaining data therefore provides the natural reference for successful deletion, but it can be expensive. Empirical machine unlearning instead aims to approximate retraining at substantially lower cost.

In image classification, most evaluations assess unlearning immediately after the model has been updated. These evaluations range from standard utility and forgetting metrics [8, 6] to more targeted comparisons of predictive behavior [7, 4], but largely evaluate the model at a single point in time. An unlearned model may later be trained on new data, adapted to downstream tasks, updated in response to further deletion requests, or exposed to relearning attacks [15, 32, 27]. Differences that are invisible immediately after unlearning may become consequential during such further training. This leads to our central question:

When an unlearned model matches retraining in its predictions, what else must it recover for that agreement to persist under further training; and do common unlearning algorithms achieve this?

We argue that a key part of the answer is the feature solution selected by retraining: which predictive features a model learns and how strongly it represents them. Two models can make similar predictions using different features and therefore respond differently when training continues.

We first ask when differences in learned features can remain harmless. Different representations can produce the same predictions under a fixed head, but this agreement may disappear once the head is updated. We show that preserving agreement across all one-step head updates requires matching model-defined feature directions and strengths. Feature recovery therefore affects whether agreement with retraining persists during adaptation.

RetrainInitial modelIUReprUnlLTDSalUnFTNegGrad+AMUNRL
Feature Discrepancy0.0000.0490.0500.0790.0990.1340.1390.1440.1510.160
Mean Accuracy (%)90.2989.7189.6985.6185.4980.4079.6284.4478.0075.46

Table 1. Missing retraining features predict downstream accuracy loss. ViT-Tiny/CIFAR-100 with 20% of the training data removed; a fresh linear head is trained on Flowers-102 on top of each frozen representation. Feature discrepancy measures how much of the retraining feature solution is missing on Flowers-102 inputs (§ 5). Retraining transfers as well as the full-data model, while unlearning methods lose accuracy as their discrepancy grows.

Next, we show that unlearning can require learning new features. One might hope that retraining after deletion largely preserves the features learned on the full data, so that successful unlearning only needs to remove or weaken existing information. We show otherwise: in a simple theoretical construction, the full model relies on an easy shortcut, while deleting an arbitrarily small fraction of the data causes retraining to suppress it and learn a predictive feature that the full model never had. Matching retraining can therefore require learning a new feature, rather than merely erasing old ones. We also show that several approximate-unlearning dynamics either fail to produce this transition or can require optimization time approaching that of retraining.

Empirically, we distinguish feature change from feature recovery. Using the model-defined feature framework of Jiang et al. [18], we introduce metrics for comparing unlearned and retrained feature solutions. These separately capture missing features that appear during retraining and features that should disappear after unlearning. Together with comparison to the full-data solution, this distinguishes feature persistence from off-target change. On ViT-Tiny/CIFAR-100, we evaluate eight approximate unlearning methods for image classification and find that methods with strong static unlearning performance can nevertheless produce very different future solutions: some remain close to the full-data model, while others move far from it without approaching retraining. We observe the same qualitative behavior with ResNet-56, SVHN, and smaller deletion fractions.

Finally, we connect feature mismatch to behavior under adaptation. Feature mismatch on the forget distribution tracks how far the model deviates from retraining after partial relearning. On downstream tasks, missing retraining features on target inputs similarly track losses in transfer performance (Table 1). We then test these relationships more directly by intervening on the feature components identified in our comparison: we restore missing retraining features and remove unmatched unlearned features. Reducing the measured mismatch through these interventions generally brings the model closer to retraining. Together, these results motivate evaluating unlearning not only by immediate behavior, but also by whether it recovers the feature solution needed to remain aligned with retraining under subsequent adaptation.

Related Work

Representation-level unlearning. Recent work goes beyond output-level evaluation. Almudévar & Ortega [1] and Sepahvand et al. [26] aim to suppress forget-set information in learned representations. Others show that output-level forgetting can leave representations close to the full model ([19]), retain forgotten-class information ([9]), or differ from retraining despite output agreement ([33]). We distinguish missing retraining features from unmatched features in the unlearning model. In particular, we ask which features retraining selects, whether unlearning algorithms recover them, and whether missing them affects downstream adaptation. Our analysis builds on the model-defined feature framework of Jiang et al. [18]; we discuss connections to representation-similarity measures in Appendix F.

Post-unlearning adaptation and relearning. Prior work studies whether forgotten behavior can re-emerge under subsequent optimization ([15, 32, 27])or whether apparently forgotten class information remains recoverable from learned representations ([3, 16]). Siddiqui et al. [27], in particular, relates relearning after retain-set fine-tuning to weight-space displacement from the full-data model. We instead ask whether unlearning recovers the features selected by retraining and whether failure to do so leads to divergence from retraining under subsequent adaptation, theoretically under head adaptation and empirically under relearning and downstream transfer.

Additional work on unlearning guarantees, methods, and evaluation is discussed in Appendix § 9.

Preliminaries

Machine Unlearning Setup. Let D=Dr⊔DfD = D_r \sqcup D_f be the training set, partitioned into a retain set DrD_r and a forget set DfD_f. For a possibly randomized training algorithm A\mathcal{A}, let

θfull∼A(D),θretr∼A(Dr),θunl∼U(θfull,Df,Dr),\theta_{\mathrm{full}} \sim\mathcal{A}(D), \qquad\theta_{\mathrm{retr}} \sim\mathcal{A}(D_r), \qquad\theta_{\mathrm{unl}} \sim\mathcal{U}(\theta_{\mathrm{full}}, D_f, D_r),

where U\mathcal{U} is an unlearning method. The ideal unlearning target is to match the distribution of θretr\theta_{\mathrm{retr}}. In practice, approximate unlearning is often evaluated by the weaker criterion of output agreement with retraining. We study whether this agreement extends to the feature solution selected by retraining, and whether feature-level differences become observable under subsequent model use.

Model Representations and Features. We follow the model-defined feature framework of Jiang et al. [18]. For a model fθ(x)=Wθ⊤ϕθ(x)f_\theta(x) = W_\theta^\top\phi_\theta(x) and a comparison set X={x1,…,xN}\mathcal{X} = \{x_1, \ldots, x_N\}, let Φθ=[ϕθ(x1),…,ϕθ(xN)]⊤\Phi_\theta= [\phi_\theta(x_1), \ldots, \phi_\theta(x_N)]^\top be the representation matrix, and let Φ~θ=(IN−1N11⊤)Φθ\widetilde{\Phi}_\theta= (I_N - \frac{1}{N}\mathbf{1}\mathbf{1}^\top)\Phi_\theta be its column-centered version. Write its Singular Value Decomposition as

Φ~θ=UθΣθVθ⊤,Uθ=[uθ,1,uθ,2,…],λθ,i=(Σθ)ii2.\widetilde{\Phi}_\theta= U_\theta\Sigma_\theta V_\theta^\top, \qquad U_\theta= [u_{\theta,1}, u_{\theta,2}, \ldots], \qquad\lambda_{\theta,i} = (\Sigma_\theta)_{ii}^{2}.

The left singular vector uθ,iu_{\theta,i} gives the normalized response pattern of the iith PCA direction across X\mathcal{X}, with strength λθ,i\lambda_{\theta,i}.

Let R\mathcal{R} be a set of independently trained models obtained under the same training setup. From each model θ∈R\theta\in\mathcal{R}, we take the top-KK PCA response patterns and cluster them jointly by similarity, yielding clusters C1,…,CJRC_1, \ldots, C_{J_{\mathcal{R}}}. Each cluster CtC_t defines a feature, represented by its sign-aligned response centroid ctc_t and average strength wt=∣Ct∣−1∑(θ,i)∈Ctλθ,iw_t = |C_t|^{-1} \sum_{(\theta,i)\in C_t} \lambda_{\theta,i}. We collect these strengths in wR=(w1,…,wJR)⊤w_{\mathcal{R}} = (w_1, \ldots, w_{J_{\mathcal{R}}})^\top. Details of the clustering and centroid construction are given in Appendix § 10.

Why Feature Mismatch Matters Under Adaptation

Empirical unlearning evaluations are typically static: they compare the unlearned and retrained models immediately after unlearning. In practice, however, an unlearned model may later be fine-tuned, trained on additional data, or adapted to a downstream task. We ask when predictive agreement between unlearned and retrained models remains stable under such updates.

We begin with a lightweight adaptation regime in which the learned representation is frozen and only the prediction head is updated. Such updates arise naturally in relearning attacks and in downstream adaptation with a frozen pretrained backbone. Proofs of all results in this section are in Appendix § 11.

Consider retrained and unlearned models with fretr(x),funl(x)∈RCf_{\mathrm{retr}}(x), f_{\mathrm{unl}}(x) \in\mathbb{R}^{C},

fretr(x)=Wretr⊤ϕretr(x),funl(x)=Wunl⊤ϕunl(x).f_{\mathrm{retr}}(x) = W_{\mathrm{retr}}^\top\phi_{\mathrm{retr}}(x), \qquad f_{\mathrm{unl}}(x) = W_{\mathrm{unl}}^\top\phi_{\mathrm{unl}}(x).

Fix a labeled dataset {(xi,yi)}i=1N\{(x_i,y_i)\}_{i=1}^{N} with input set X={x1,…,xN}\mathcal{X} = \{x_1,\ldots,x_N\}, and suppose the models are initially predictively equivalent on X\mathcal{X},

Wretr⊤ϕretr(x)=Wunl⊤ϕunl(x)∀x∈X.(1)W_{\mathrm{retr}}^\top\phi_{\mathrm{retr}}(x) = W_{\mathrm{unl}}^\top\phi_{\mathrm{unl}}(x) \qquad\forall x \in\mathcal{X}. \tag*{(1)}

For a differentiable loss ℓ(z,y)\ell(z,y) applied to the prediction z=f(x)z = f(x), define for each (xi,yi)(x_i,y_i)

gi:=∇zℓ(z,yi)∣z=fretr(xi)=∇zℓ(z,yi)∣z=funl(xi)∈RC,g_i := \left.\nabla_z \ell(z,y_i)\right|_{z=f_{\mathrm{retr}}(x_i)} = \left.\nabla_z \ell(z,y_i)\right|_{z=f_{\mathrm{unl}}(x_i)} \in\mathbb{R}^{C},

where the equality follows from predictive equivalence. Updating only the head on (xi,yi)(x_i,y_i) with step size η\eta gives

Wm,i+=Wm−ηϕm(xi)gi⊤,m∈{retr,unl}.W_{m,i}^{+} = W_m - \eta\phi_m(x_i)g_i^\top, \qquad m \in\{\mathrm{retr},\mathrm{unl}\}.

Let Fm,i+∈RN×CF_{m,i}^{+} \in\mathbb{R}^{N \times C} collect the predictions on X\mathcal{X} after this update, and let Φm\Phi_m and Φ~m\widetilde{\Phi}_m denote the uncentered and centered representation matrices on X\mathcal{X}, respectively.

The result below shows that predictive equivalence can be fragile under even a single head update.

Proposition 3.1. Suppose that gi≠0g_i \ne0 for all ii and that fretr(x)=funl(x)f_{\mathrm{retr}}(x) = f_{\mathrm{unl}}(x) for all x∈Xx \in\mathcal{X}. Then

∑i=1N∥Fretr,i+−Funl,i+∥F2η2∥gi∥22=∥ΦretrΦretrT−ΦunlΦunlT∥F2≥∥Φ~retrΦ~retrT−Φ~unlΦ~unlT∥F2.\sum_{i=1}^{N} \frac{\left\|F_{\mathrm{retr},i}^{+} - F_{\mathrm{unl},i}^{+}\right\|_{F}^{2}}{\eta^{2}\left\|g_i\right\|_{2}^{2}} = \left\|\Phi_{\mathrm{retr}}\Phi_{\mathrm{retr}}^{\mathsf{T}} - \Phi_{\mathrm{unl}}\Phi_{\mathrm{unl}}^{\mathsf{T}}\right\|_{F}^{2} \ge\left\|\widetilde{\Phi}_{\mathrm{retr}}\widetilde{\Phi}_{\mathrm{retr}}^{\mathsf{T}} - \widetilde{\Phi}_{\mathrm{unl}}\widetilde{\Phi}_{\mathrm{unl}}^{\mathsf{T}}\right\|_{F}^{2}.

The proposition quantifies how representation mismatch can be exposed by even a single head update. The centered Gram matrix decomposes as Φ~Φ~T=∑jλjujujT\widetilde{\Phi}\widetilde{\Phi}^{\mathsf{T}} = \sum_j \lambda_j u_j u_j^{\mathsf{T}}, so discrepancies in it reflect differences in both the response directions and strengths of model-defined features. As a special case, if output equivalence on X\mathcal{X} is preserved after updating on each (xi,yi)(x_i, y_i), i∈[N]i \in[N], the centered Gram matrices must coincide, and the two models therefore have the same feature structure on X\mathcal{X}, up to the usual eigenspace ambiguities.

More generally, larger centered Gram mismatch implies larger normalized prediction disagreement after adaptation. The bound is one-sided: small centered mismatch does not rule out additional disagreement from components removed by centering. When the adaptation gradient is non-negligible, this lower bound translates centered Gram mismatch into a non-negligible prediction gap, which is particularly relevant for relearning on the forget set.

A fixed-head perspective. The analysis above considers output equivalence between complete retrained and unlearned models. Alternatively, if unlearning is viewed as removing forget-set information from the model, the representation and prediction head may be judged separately rather than required to jointly match the same retrained model. Under this view, a retrained representation ϕretr\phi_{\mathrm{retr}} is indistinguishable from ϕunl\phi_{\mathrm{unl}} under the current unlearned head if

WunlT(ϕretr(x)−ϕunl(x))=0∀x∈X.W_{\mathrm{unl}}^{\mathsf{T}}\left(\phi_{\mathrm{retr}}(x) - \phi_{\mathrm{unl}}(x)\right) = 0 \qquad\forall x \in\mathcal{X}.

Such a difference is “harmless” in the sense of Grant et al. [12]. In Appendix § 11.§ 2, we show that this harmlessness is generally unstable: once the head is adapted, previously invisible representation differences can affect the predictions.

Data Deletion Can Require Learning New Features

References

  1. [1]Antonio Almudévar and Alfonso Ortega. Representation unlearning: Forgetting through information compression. In Forty-third International Conference on Machine Learning, 2026. URL https://openreview.net/forum?id=CWjK5eUzCI.arxiv.org/abs/2601.21564
  2. [2]Vikram S Chundawat, Ayush K Tarun, Murari Mandal, and Mohan Kankanhalli. Can bad teaching induce forgetting? unlearning in deep networks using an incompetent teacher. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp. 7210–7217, 2023.
  3. [3]Zahra Dehghani, Pablo Piantanida, and Mohammad hadi Shateri. Source-free class relearning: Diagnosing forgetting in class unlearning. arXiv preprint arXiv:2609.02018, 2026.arxiv.org/abs/2609.02018
  4. [4]Polina Dolgova and Sebastian U. Stich. Forgetting has neighbors: Localized collateral forgetting in machine unlearning, 2026. URL https://arxiv.org/abs/2605.31317.
  5. [5]Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale. CoRR, abs/2010.11929, 2020. URL https://arxiv.org/abs/2010.11929.
  6. [6]Ali Ebrahimpour-Boroojeny, Hari Sundaram, and Varun Chandrasekaran. Amun: Adversarial machine unlearning, 2025a. URL https://arxiv.org/abs/2503.00917.
  7. [7]Ali Ebrahimpour-Boroojeny, Yian Wang, and Hari Sundaram. On the necessity of output distribution reweighting for effective class unlearning, 2025b. URL https://arxiv.org/abs/2506.20893.
  8. [8]Chongyu Fan, Jiancheng Liu, Yihua Zhang, Eric Wong, Dennis Wei, and Sijia Liu. Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=gn0mThQGNM.arxiv.org/abs/2310.12508
  9. [9]Yichen Gao, Altay Unal, Akshay Rangamani, and Zhihui Zhu. An illusion of unlearning? assessing machine unlearning through internal representations. arXiv preprint arXiv:2604.08271, 2026.arxiv.org/abs/2604.08271
  10. [10]Shashwat Goel, Ameya Prabhu, Amartya Sanyal, Ser-Nam Lim, Philip Torr, and Ponnurangam Kumaraguru. Towards adversarial evaluations for inexact machine unlearning. arXiv preprint arXiv:2201.06640, 2022.arxiv.org/abs/2201.06640
  11. [11]Shashwat Goel, Ameya Prabhu, Philip Torr, Ponnurangam Kumaraguru, and Amartya Sanyal. Corrective machine unlearning. arXiv preprint arXiv:2402.14015, 2024.arxiv.org/abs/2402.14015
  12. [12]Satchel Grant, Simon Jerome Han, Alexa R. Tartaglini, and Christopher Potts. Addressing divergent representations from causal interventions on neural networks. In The Fourteenth International Conference on Learning Representations, 2026. URL https://openreview.net/forum?id=cZrTMqYVL6.arxiv.org/abs/2511.04638
  13. [13]Jamie Hayes, Ilia Shumailov, Eleni Triantafillou, Amr Khalifa, and Nicolas Papernot. Inexact unlearning needs more careful evaluations to avoid a false sense of privacy. In 2025 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), pp. 497–519. IEEE, 2025.arxiv.org/abs/2403.01218
  14. [14]Kaiming He, X. Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2015. URL https://api.semanticscholar.org/CorpusID:206594692.arxiv.org/abs/1512.03385
  15. [15]Shengyuan Hu, Yiwei Fu, Steven Wu, and Virginia Smith. Unlearning or obfuscating? jogging the memory of unlearned LLMs via benign relearning. In The Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=fMNRyBvcQN.arxiv.org/abs/2406.13356
  16. [16]Yurim Jang, Jaeung Lee, Dohyun Kim, Jaemin Jo, and Simon S Woo. Suppression or deletion: A restoration-based representation-level analysis of machine unlearning. In Proceedings of the ACM Web Conference 2026, pp. 8505–8508, 2026.arxiv.org/abs/2602.18505
  17. [17]Jinghan Jia, Jiancheng Liu, Parikshit Ram, Yuguang Yao, Gaowen Liu, Yang Liu, Pranay Sharma, and Sijia Liu. Model sparsity can simplify machine unlearning. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=0jZh883i34.arxiv.org/abs/2304.04934
  18. [18]Yiding Jiang, Christina Baek, and J Zico Kolter. On the joint interaction of models, data, and features. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=ze7D0Li394.arxiv.org/abs/2306.04793
  19. [19]Yongwoo Kim, Sungmin Cha, and Donghyun Kim. Are we truly forgetting? a critical examination of machine unlearning evaluation protocols. Engineering Applications of Artificial Intelligence, 167:113785, 2026.DOI
  20. [20]Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions. In Doina Precup and Yee Whye Teh (eds.), Proceedings of the 34th International Conference on Machine Learning, volume 70 of Proceedings of Machine Learning Research, pp. 1885–1894. PMLR, 06–11 Aug 2017. URL https://proceedings.mlr.press/v70/koh17a.html.arxiv.org/abs/1703.04730
  21. [21]Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. Technical Report 0, University of Toronto, Toronto, Ontario, 2009. URL https://www.cs.toronto.edu/~kriz/learning-features-2009-TR.pdf.
  22. [22]Meghdad Kurmanji, Peter Triantafillou, Jamie Hayes, and Eleni Triantafillou. Towards unbounded machine unlearning. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=OveBaTtUAT.
  23. [23]Seth Neel, Aaron Roth, and Saeed Sharifi-Malvajerdi. Descent-to-delete: Gradient-based methods for machine unlearning. In Vitaly Feldman, Katrina Ligett, and Sivan Sabato (eds.), Proceedings of the 32nd International Conference on Algorithmic Learning Theory, volume 132 of Proceedings of Machine Learning Research, pp. 931–962. PMLR, 16–19 Mar 2021. URL https://proceedings.mlr.press/v132/neel21a.html.arxiv.org/abs/2007.02923
  24. [24]Maria-Elena Nilsback and Andrew Zisserman. Automated flower classification over a large number of classes. In Indian Conference on Computer Vision, Graphics and Image Processing, Dec 2008.DOI
  25. [25]Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman, and C. V. Jawahar. Cats and dogs. In IEEE Conference on Computer Vision and Pattern Recognition, 2012.
  26. [26]Nazanin Mohammadi Sepahvand, Eleni Triantafillou, Hugo Larochelle, Doina Precup, James J. Clark, Daniel M. Roy, and Gintare Karolina Dziugaite. Selective unlearning via representation erasure using domain adversarial training. In The Thirteenth International Conference on Learning Representations, 2025. URL https://openreview.net/forum?id=KzSGJy1PIf.
  27. [27]Shoaib Ahmed Siddiqui, Adrian Weller, David Krueger, Gintare Karolina Dziugaite, Michael Curtis Mozer, and Eleni Triantafillou. From dormant to deleted: Tamper-resistant unlearning through weight-space regularization. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025. URL https://openreview.net/forum?id=Zrqn7ZshXG.arxiv.org/abs/2505.22310
  28. [28]Anyith Thudi, Gabriel Deza, Varun Chandrasekaran, and Nicolas Papernot. Unrolling SGD: Understanding Factors Influencing Machine Unlearning . In 2022 IEEE 7th European Symposium on Security and Privacy (EuroS&P), pp. 303–319, Los Alamitos, CA, USA, June 2022. IEEE Computer Society. doi: 10.1109/EuroSP53844.2022.00027. URL https://doi.ieeecomputersociety.org/10.1109/EuroSP53844.2022.00027.DOI
  29. [29]Yiwen Tu, Pingbang Hu, and Jiaq Ma. A reliable cryptographic framework for empirical machine unlearning evaluation. Advances in Neural Information Processing Systems, 38:131705–131736, 2026.arxiv.org/abs/2404.11577
  30. [30]Alexander Warnecke, Lukas Pirch, Christian Wressnegger, and Konrad Rieck. Machine unlearning of features and labels. In Network and Distributed System Security (NDSS) Symposium 2023: 27 February - 3 March 2023, San Diego, CA, USA, 2023. ISBN 1-891562-83-5. doi: 10.14722/ndss.2023.23087. 46.23.01; LK 01.
  31. [31]Heng Xu, Tianqing Zhu, Lefeng Zhang, Wanlei Zhou, and Philip S. Yu. Machine unlearning: A survey, 2023. URL https://arxiv.org/abs/2306.03558.
  32. [32]Xiaoyu Xu, Xiang Yue, Yang Liu, Qingqing Ye, Huadi Zheng, Peizhao Hu, Minxin Du, and Haibo Hu. Unlearning isn’t deletion: Investigating reversibility of machine unlearning in LLMs. In Forty-third International Conference on Machine Learning, 2026. URL https://openreview.net/forum?id=E5SVwOI3b.arxiv.org/abs/2505.16831
  33. [33]Teresa Pui Yee Yong, Win Kent Ong, and Chee Seng Chan. Erased, but not gone: Output forgetting is not true forgetting, 2026. URL https://arxiv.org/abs/2606.25001.
  34. [34]Vladimir Zorich. Mathematical analysis. Part I. 01 2016. ISBN 978-3-662-48991-8. doi: 10.1007/978-3-662-48993-2.

Paper details

vlnc:W_df5mfuq2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents