Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Speech Unlearning

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper defines speech unlearning — erasing one recording or an entire speaker from a trained model without retraining — and reports that all five standard methods fail: each either leaves the targeted data recognizable or destroys…

desk verdict First speech-unlearning benchmark, but the headline cross-modal claim outruns the experiments. read the letter →

arxiv 2506.00848 v1 pith:3HVP7YMG submitted 2025-06-01 cs.LG cs.AIcs.SDeess.AS

Jiali Cheng , Hadi Amiri This is my paper
classification cs.LGcs.AIcs.SDeess.AS
keywords speechunlearningmachinekeywordspottingspeakeridentificationmembershipinferenceattackrighttobeforgottencurriculumlearningsamplereweighting
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

Speech unlearning is the task of removing a chosen voice recording (sample unlearning) or an entire speaker or keyword (class unlearning) from an already trained speech model without retraining from scratch. The paper argues this matters for privacy law compliance, for deleting outdated or sensitive content such as wake words, and for correcting biased or noisy training data. Its central empirical claim is negative: across keyword spotting and speaker identification, every one of five standard unlearning methods adopted from vision and text research either leaves the targeted data recognizable or destroys accuracy on retained and test data. The paper's positive result is that a structured forgetting objective based on curriculum-style sample reweighting (SuperLoss) cuts the forget-set score from 44.7 to 17.3, pointing to a research direction rather than a finished solution.

What carries the argument

The central machinery is a unified unlearning objective, $f' = \arg\min_{\theta'} \mathcal{L}_f(\mathcal{D}_f) + \lambda \mathcal{L}_r(\mathcal{D}_r)$, which casts speech unlearning as fine-tuning the original model to minimize a forgetting loss on the forget set and a retention loss on the retain set; each tested method is one choice of these two terms, and $\lambda$ sets the trade-off. The two tasks are sample unlearning (erase one recording) and class unlearning (erase a speaker or keyword), evaluated by accuracy on the forget, retain, and original test sets plus a membership inference attack on the forget set. The paper's own promising result runs through SuperLoss, $\mathcal{L}_\lambda = (l_i - \tau)\sigma_i + \lambda(\log\sigma_i)^2$, where $\tau$ is a moving-average loss threshold and each sample weight $\sigma_i$ is given in closed form; this turns flat fine-tuning into a curriculum that reweights hard-to-forget samples. Experiments average across seven speech backbones — Whisper Tiny and Base, Wav2Vec 2.0 Base and Large, HuBERT Base, Large, and X-Large — on the Speech Commands and VoxCeleb1 datasets.

What would settle it

Re-measure membership inference as a plain attack-success rate: the fraction of forget-set samples that an attack model labels as training members, reported for the original model and each unlearned model. The original model trained on the forget set, so it should show the highest success rate; whichever quantity in Tables 1 and 2 lines up with that fact settles the metric's direction, and with it whether the SuperLoss reduction from 44.7 to 17.3 is an improvement or a return to original-level memorization.

Watch

Extended reading notes

Core claim

The paper claims that speech unlearning is genuinely harder than unlearning in vision or text, and that off-the-shelf unlearning methods fail on it. On keyword spotting and speaker identification, five methods — gradient ascent, random labeling, saliency-based unlearning (SalUn), SCRUB, and bad-teacher training — are tested across Whisper, Wav2Vec 2.0, and HuBERT backbones. The failure pattern is consistent: gradient ascent, SCRUB, and bad-teacher forget the target data but collapse accuracy on everything else, while random labeling and SalUn preserve accuracy but leave forget-set samples recognizable, as judged by subset accuracy and by a black-box membership inference attack. Class unlearning, removing an entire keyword or speaker, proves harder than sample unlearning. The paper also reports that a structured forgetting objective using SuperLoss — a curriculum-style loss that reweights each sample by a closed-form function of its own loss — improves the forget-set score from 44.7 to 17.3, and frames speech unlearning as an open area with feature-level, certifiable, and adversarial directions to explore.

Load-bearing premise

The conclusions depend on one reading of the membership-inference score (a test of whether the model still recognizes supposedly erased recordings): the paper treats a higher score as meaning the erased data stays detectable, yet the original trained model reports the lowest score in both tables, so the direction of this metric is the load-bearing assumption.

Editorial extensions

If this is right

  • Unlearning methods built for images and text do not transfer to speech; the uniform failure pattern in Tables 1 and 2 means the forgetting and retention objectives have to be redesigned for temporal, speaker-entangled representations.
  • Class unlearning is the harder task: in every method, erasing an entire keyword or speaker drags down recognition of the remaining classes more than sample unlearning does.
  • Forgetting-first methods (gradient ascent, SCRUB, bad teacher) and retention-first methods (random labeling, SalUn) fail in opposite, predictable ways, so a working method needs both mechanisms rather than either extreme.
  • The SuperLoss experiment, cutting the forget-set score from 44.7 to 17.3, indicates that how samples are weighted and scheduled during unlearning matters as much as the gradient update itself.
  • If speech unlearning works, concrete applications open up: retiring wake words, removing a deactivated user's voice from authentication systems, deleting outdated commands, and de-identifying speakers in medical or legal speech records.

Reading between the lines

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

  • My inference: the membership-inference numbers in the paper are internally ambiguous — the original model that trained on the forget set has the lowest score in both tables, while every unlearned model scores higher, and the metric's definition in section 4.4 and its use in section 5.3 point in opposite directions; the metric direction needs one explicit statement to decide which reading is right.
  • A testable extension the paper does not run: order the forget set by per-sample loss and vary the SuperLoss threshold schedule, to see whether class unlearning improves monotonically with curriculum strength; the closed-form weights make this a minor change to the reported setup.
  • If the entanglement claim is right, feature-level unlearning — removing pitch, timbre, or accent while keeping linguistic content — is not a niche add-on but the core of the problem: successful speech unlearning must first separate speaker identity from phonetic content in the representation space.
  • A natural next benchmark is a relearning attack: after supposedly forgetting a speaker, fine-tune the unlearned model briefly and measure how many steps re-identify that speaker; the paper cites such attacks but does not run them, and they would quantify exactly how much residual knowledge each failure mode leaves.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper introduces machine unlearning for speech, defining sample unlearning (removing individual recordings) and class unlearning (removing a speaker or keyword class). It casts unlearning as an optimization problem in Eq. (1) with forgetting and retention losses, and it evaluates five existing methods (Gradient Ascent, Random Labeling, SalUn, SCRUB, Bad-T) on keyword spotting (Speech Commands) and speaker identification (VoxCeleb1), reporting accuracy on the forget, retain, and test sets, a membership inference attack (MIA) score, and runtime. The authors report that none of the tested methods successfully unlearns speech data without degrading retained accuracy, and they claim in the abstract and in §5.3 that speech unlearning is 'significantly more challenging than unlearning image or text data.' They also propose a SuperLoss-based structured forgetting strategy that they report reduces a forget-set score from 44.7 to 17.3, and they outline future directions.

Significance. If the empirical findings are properly supported, this would be a useful first systematic study of unlearning for speech, with clear task definitions, two complementary speech tasks, several baseline methods, and a time-constrained training protocol. The authors are honest about the limitations of current methods, and the observation that label-randomization and saliency methods retain substantial knowledge of the forget set while gradient-ascent methods destroy retain-set utility is a concrete, falsifiable finding. However, the paper's headline comparative claim (speech is harder than image/text), the interpretation of the MIA metric, and the reported SuperLoss improvement are not currently supported by the evidence as presented. The paper's contribution is better framed as a benchmark and difficulty analysis of existing unlearning methods on speech, rather than a demonstration that speech is inherently harder than other modalities.

major comments (4)
  1. [Abstract and §5.3] The abstract states that 'unlearning speech data is significantly more challenging than unlearning image or text data,' but no image or text unlearning baseline is run in §5, and no matched comparison to published vision/NLP unlearning results is provided. This is a load-bearing comparative claim about cross-modal difficulty, and without controlled experiments or a clearly justified comparison protocol it is not supported. The authors should either remove the cross-modal claim or add matched image/text unlearning experiments under the same evaluation framework.
  2. [§4.4 and Tables 1-2] The MIA metric is internally inconsistent as reported. In §4.4 the text says that if an MIA model classifies most Df samples as non-members, this indicates the unlearned model has limited knowledge of Df, which suggests the reported MIA(↑) is the rate of correct non-member classification. Under that reading, the original model's low MIA of 17.3 is sensible (it has trained on Df and therefore classifies those samples as members), but the unlearned models' MIA values around 50 percent indicate chance-level membership discrimination, which would mean the samples are not clearly detectable. In contrast, §5.3 interprets higher MIA as 'unlearned samples remain detectable,' which would imply that the original model is the least detectable, an implausible outcome. The direction of the metric and the threshold for 'detectable' must be defined precisely, and the conclusion in §5.3 must be re-derived from the chosen definition.
  3. [§5.4] The SuperLoss result is not verifiable from the manuscript. The text says 'the overall unlearning performance on Df effectively decreased from 44.7 to 17.3,' but no SuperLoss row appears in Tables 1 or 2, and the baseline value 44.7 coincides with the MIA value for RandomLabel class unlearning in Table 1 while 17.3 is the original model's MIA in the same table. The authors do not state which metric (Df accuracy, MIA, or another score) is being reported, which method SuperLoss is applied to, or under which setup. Since this is presented as a promising solution, full experimental details and a table row or figure are required.
  4. [§5.2 and Tables 1-2] The reported results are aggregated averages across several backbone architectures (Whisper, Wav2Vec 2.0, HuBERT) and across forget-set fractions from 1% to 10%, but the tables contain only point estimates with no standard deviations, no per-architecture breakdown, and no description of how the averaging is performed. The claim that all methods fail is probably robust, but specific quantitative comparisons (for example, Df values of 12.3 vs. 11.2 for GradAscent and SCRUB) cannot be evaluated without variance information. The authors state that all experiments were repeated five times, so these standard deviations should be reported.
minor comments (6)
  1. [Figure 1] The figure caption contains a typo: 'Utterence' should be 'Utterance.'
  2. [§5.3] There is a typo in the first paragraph: 'suggeting' should be 'suggesting.'
  3. [§7] The conclusion contains a duplicated article: 'the the high-dimensional, sequential, and speaker-dependent nature' should read 'the high-dimensional, sequential, and speaker-dependent nature.'
  4. [§5.1] The dataset name is typeset as 'V oxCeleb1' with an extra space; it should be 'VoxCeleb1.'
  5. [§4.4] The phrase 'If a MIA model classifies most samples of Df as non-members' should specify that the subject is a membership-inference classifier, not a model, to avoid confusion with the speech model being evaluated.
  6. [§4.2 and Table 1] The authors do not state the random-chance accuracy levels for the forget set. For speaker identification with 1,211 speakers, an accuracy of 33.4 is far above chance, while for keyword spotting with 12 classes the interpretation is different; stating chance levels would help readers interpret Df values.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical benchmark results with no derivation that reduces to its own inputs.

full rationale

This paper is an empirical benchmark study, not a derivation. The central results are accuracy and MIA scores in Tables 1 and 2, obtained by applying standard unlearning methods (§4.3, Eq. 1) to speech datasets and measuring downstream performance. No parameter is fitted to the target outcome: the SuperLoss-based structured forgetting result (§5.4) is an experimental application of an existing loss (SuperLoss, [38]) and is not derived from the Df scores it reports. The unified optimization formulation in Eq. (1) is a standard framework supported by multiple independent prior works ([3, 29, 30]), so the self-citations [19, 20, 22, 25, 27, 28] are contextual related-work citations and are not load-bearing for any conclusion. The abstract's claim that speech unlearning is 'significantly more challenging than unlearning image or text data' is a comparative claim without matched cross-modal baselines, but this is an evidential/correctness gap, not circularity: the claim is not obtained by assuming itself. The MIA-direction ambiguity noted by the reader is a metric interpretation issue, not a circular reduction. Overall, the paper's empirical conclusions are self-contained with respect to its own measurements, so no circularity is present.

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

The paper introduces no new entities or derivations. Its evidence rests on empirical benchmark results whose hyperparameters and metric definitions are under-specified. The only free parameters are the loss weights and thresholds that are never reported.

free parameters (3)
  • trade-off weight λ in Eq. (1)
    Controls the balance between forgetting and retention for each method; values are not reported for any experiment.
  • SuperLoss hyperparameters λ and τ
    τ is the moving-average loss threshold and λ is the regularization weight; neither is reported for the structured forgetting experiment.
  • MIA decision threshold
    Black-box MIA requires a threshold on loss values to decide membership; the threshold is not specified.
assumptions (4)
  • domain assumption Accuracy on the forget set Df is a valid proxy for successful unlearning.
    Used throughout §5 as the primary forgetting metric, but accuracy drops could come from suppression rather than true removal of influence.
  • domain assumption The MIA attack is correctly implemented and its scores are interpreted consistently.
    The reported MIA values for the original model suggest the metric may measure the non-member fraction, which conflicts with the prose interpretation in §5.3.
  • domain assumption Whisper, Wav2Vec 2.0, and HuBERT are representative speech model architectures.
    Results are averaged across these models and treated as generalizable to speech models, but no diversity analysis is given.
  • domain assumption Existing unlearning methods were applied with reasonable hyperparameter settings.
    No hyperparameters are reported, so the difficulty conclusion assumes the methods were not under-tuned or poorly configured.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Speech Unlearning." pith.science (2026). https://pith.science/paper/3HVP7YMG

@misc{pith2026250600848,
  author       = {Pith},
  title        = {Pith review of: Speech Unlearning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3HVP7YMG}},
  note         = {Machine review of arXiv:2506.00848}
}
read the original abstract

We introduce machine unlearning for speech tasks, a novel and underexplored research problem that aims to efficiently and effectively remove the influence of specific data from trained speech models without full retraining. This has important applications in privacy preservation, removal of outdated or noisy data, and bias mitigation. While machine unlearning has been studied in computer vision and natural language processing, its application to speech is largely unexplored due to the high-dimensional, sequential, and speaker-dependent nature of speech data. We define two fundamental speech unlearning tasks: sample unlearning, which removes individual data points (e.g., a voice recording), and class unlearning, which removes an entire category (e.g., all data from a speaker), while preserving performance on the remaining data. Experiments on keyword spotting and speaker identification demonstrate that unlearning speech data is significantly more challenging than unlearning image or text data. We conclude with key future directions in this area, including structured training, robust evaluation, feature-level unlearning, broader applications, scalable methods, and adversarial robustness.

Figures

Figures reproduced from arXiv: 2506.00848 by the authors.

Figure 1
Figure 1. Illustration of Speech Unlearning. • we formally define the task of speech unlearning and high￾light its importance and challenges, • we conduct experiments on keyword spotting and speaker identification to quantify the difficulty of disentangling learned speech representations and the trade-offs between un￾learning effectiveness and model utility, and • we outline open challenges and promising research directions t… view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks

    cs.LG 2026-07 conditional novelty 5.0 of 10

    A system-first taxonomy and literature synthesis of multimodal unlearning across vision, language, video, and audio, with datasets, benchmarks, metrics, applications, and open challenges.

Reference graph

Works this paper leans on

46 extracted references · 40 canonical work pages · cited by 1 Pith paper

  1. [1]

    Hey Google

    Introduction Machine unlearning [1, 2] aims to remove specific data (e.g., training examples or individual classes) from a trained model while preserving performance on the remaining data. While progress has been made in computer vision [2], natural lan- guage processing [3], and recommendation systems [4], its ap- plication to speech models is largely un...

    work page

  2. [2]

    Existing methods approach un- learning with efficient retraining by dividing data into chunks and training a separate model on each chunk [5, 6]

    Related Work Machine unlearning was initially studied on vision tasks, specif- ically, image classification [1]. Existing methods approach un- learning with efficient retraining by dividing data into chunks and training a separate model on each chunk [5, 6]. Influence- based unlearning methods derives a one-shot update with the- oretical guarantee, but us...

    arXiv 2025

  3. [3]

    In what follows, we describe some of these challenges: 3.1

    Challenges of Speech Unlearning Speech unlearning presents unique challenges beyond those in other modalities due to the multi-dimensional, sequential, and speaker-dependent nature of speech data. In what follows, we describe some of these challenges: 3.1. Temporal Dependencies and Sequence Complexity Unlike images or text data, speech data is a continuou...

    work page

  4. [4]

    Hey Google

    Machine Unlearning in Speech 4.1. Problem Definition Given a speech modelf trained on dataset Dtrain, machine un- learning aims to remove the influence of a subset Df ∈ Dtrain (the forget set) from f without full retraining, while preserving knowledge on the retain data Dr = Dtrain \ Df . We term f as the original model and f ′ as the model post-unlearnin...

    work page

  5. [5]

    Hey Siri

    Experiments 5.1. Datasets We consider two speech tasks using data from the Superb benchmark [32]: keyword spotting and speaker identification. Keyword Spotting: the goal is to detect predefined keywords, such as “Hey Siri” or “Okay Google,” in speech recordings. In case of sample unlearning, user request the removal of their spe- cific voice recordings fr...

    work page

  6. [6]

    Future Directions Speech unlearning is still in its early stages with several open challenges and research opportunities. Building on our findings and analysis, we discuss key advancements needed to develop effective speech unlearning methods, establish stronger evalua- tion metrics, extend unlearning to broader speech applications, and address possible r...

    work page

  7. [7]

    Conclusion We introduce speech unlearning, a novel and underexplored challenge in AI that focuses on selectively removing a subset of speech data and its influence from trained models, which is crucial for privacy preservation, ethical AI, and regulatory com- pliance. While machine unlearning has been investigated in do- mains such as vision, language mod...

    work page

  8. [8]

    Machine unlearn- ing,

    L. Bourtoule, V . Chandrasekaran, C. A. Choquette-Choo, H. Jia, A. Travers, B. Zhang, D. Lie, and N. Papernot, “Machine unlearn- ing,” in IEEE Symposium on Security and Privacy (SP), 2021

    work page 2021

Show all 46 references
  1. [9]

    Eternal sunshine of the spotless net: Selective forgetting in deep networks,

    A. Golatkar, A. Achille, and S. Soatto, “Eternal sunshine of the spotless net: Selective forgetting in deep networks,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020

    2020

  2. [10]

    SOUL: Unlocking the power of second-order optimization for LLM unlearning,

    J. Jia, Y . Zhang, Y . Zhang, J. Liu, B. Runwal, J. Diffender- fer, B. Kailkhura, and S. Liu, “SOUL: Unlocking the power of second-order optimization for LLM unlearning,” in Proceedings of the 2024 Conference on Empirical Methods in Natural Lan- guage Processing, Nov. 2024, pp...

    2024

  3. [11]

    Recommendation un- learning,

    C. Chen, F. Sun, M. Zhang, and B. Ding, “Recommendation un- learning,” in Proceedings of the ACM Web Conference, 2022

    2022

  4. [12]

    Safe: Machine unlearning with shard graphs,

    Y . Dukler, B. Bowman, A. Achille, A. Golatkar, A. Swaminathan, and S. Soatto, “Safe: Machine unlearning with shard graphs,” in Proceedings of the IEEE/CVF International Conference on Com- puter Vision (ICCV), October 2023, pp. 17 108–17 118

    2023

  5. [13]

    Erm-ktp: Knowledge-level machine unlearning via knowledge transfer,

    S. Lin, X. Zhang, C. Chen, X. Chen, and W. Susilo, “Erm-ktp: Knowledge-level machine unlearning via knowledge transfer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 20 147–20 155

    2023

  6. [14]

    Algorithms that approxi- mate data removal: New results and limitations,

    V . M. Suriyakumar and A. C. Wilson, “Algorithms that approxi- mate data removal: New results and limitations,” in Advances in Neural Information Processing Systems, 2022

    2022

  7. [15]

    Certified minimax unlearning with generalization rates and deletion capacity,

    J. Liu, J. Lou, Z. Qin, and K. Ren, “Certified minimax unlearning with generalization rates and deletion capacity,” in Advances in Neural Information Processing Systems, 2023

    2023

  8. [16]

    Machine unlearn- ing for image retrieval: A generative scrubbing approach,

    P.-F. Zhang, G. Bai, Z. Huang, and X.-S. Xu, “Machine unlearn- ing for image retrieval: A generative scrubbing approach,” inPro- ceedings of the 30th ACM International Conference on Multime- dia, 2022

    2022

  9. [17]

    Safe-CLIP: Removing NSFW Concepts from Vision-and- Language Models,

    S. Poppi, T. Poppi, F. Cocchi, M. Cornia, L. Baraldi, and R. Cuc- chiara, “Safe-CLIP: Removing NSFW Concepts from Vision-and- Language Models,” in Proceedings of the European Conference on Computer Vision, 2024

    2024

  10. [18]

    Unlearncanvas: A stylized image dataset to benchmark machine unlearning for diffusion models,

    Y . Zhang, Y . Zhang, Y . Yao, J. Jia, J. Liu, X. Liu, and S. Liu, “Unlearncanvas: A stylized image dataset to benchmark machine unlearning for diffusion models,” 2024

    2024

  11. [19]

    Machine unlearn- ing for image-to-image generative models,

    G. Li, H. Hsu, C.-F. Chen, and R. Marculescu, “Machine unlearn- ing for image-to-image generative models,” in The Twelfth Inter- national Conference on Learning Representations, 2024

    2024

  12. [20]

    Tofu: A task of fictitious unlearning for llms,

    P. Maini, Z. Feng, A. Schwarzschild, Z. C. Lipton, and J. Z. Kolter, “Tofu: A task of fictitious unlearning for llms,” 2024

    2024

  13. [21]

    Who’s harry potter? approximate unlearning in llms,

    R. Eldan and M. Russinovich, “Who’s harry potter? approximate unlearning in llms,” 2023

    2023

  14. [22]

    Negative preference opti- mization: From catastrophic collapse to effective unlearning,

    R. Zhang, L. Lin, Y . Bai, and S. Mei, “Negative preference opti- mization: From catastrophic collapse to effective unlearning,” in First Conference on Language Modeling, 2024

    2024

  15. [23]

    Reversing the forget-retain objectives: An efficient LLM unlearning framework from logit difference,

    J. Ji, Y . Liu, Y . Zhang, G. Liu, R. R. Kompella, S. Liu, and S. Chang, “Reversing the forget-retain objectives: An efficient LLM unlearning framework from logit difference,” in Advances in Neural Information Processing Systems, 2024

    2024

  16. [24]

    Preserving privacy through dememorization: An unlearning technique for mitigat- ing memorization risks in language models,

    A. Kassem, O. Mahmoud, and S. Saad, “Preserving privacy through dememorization: An unlearning technique for mitigat- ing memorization risks in language models,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023

    2023

  17. [25]

    Efficient model updates for approximate unlearning of graph-structured data,

    E. Chien, C. Pan, and O. Milenkovic, “Efficient model updates for approximate unlearning of graph-structured data,” in Interna- tional Conference on Learning Representations, 2023

    2023

  18. [26]

    GN- NDelete: A general strategy for unlearning in graph neural net- works,

    J. Cheng, G. Dasoulas, H. He, C. Agarwal, and M. Zitnik, “GN- NDelete: A general strategy for unlearning in graph neural net- works,” in The Eleventh International Conference on Learning Representations, 2023

    2023

  19. [27]

    Multidelete for multimodal machine unlearning,

    J. Cheng and H. Amiri, “Multidelete for multimodal machine unlearning,” in Computer Vision – ECCV 2024 , A. Leonardis, E. Ricci, S. Roth, O. Russakovsky, T. Sattler, and G. Varol, Eds. Cham: Springer Nature Switzerland, 2025, pp. 165–184

    2024

  20. [28]

    Deep regression unlearning,

    A. K. Tarun, V . S. Chundawat, M. Mandal, and M. Kankanhalli, “Deep regression unlearning,” inProceedings of the 40th Interna- tional Conference on Machine Learning, 2023

    2023

  21. [29]

    Understanding machine unlearn- ing through the lens of mode connectivity,

    J. Cheng and H. Amiri, “Understanding machine unlearn- ing through the lens of mode connectivity,” arXiv preprint arXiv:2504.06407, 2025

    2025 arXiv

  22. [30]

    Ma- chine unlearning of federated clusters,

    C. Pan, J. Sima, S. Prakash, V . Rana, and O. Milenkovic, “Ma- chine unlearning of federated clusters,” in The Eleventh Interna- tional Conference on Learning Representations, 2023

    2023

  23. [31]

    Continual learning and private unlearning,

    B. Liu, Q. Liu, and P. Stone, “Continual learning and private unlearning,” in Proceedings of The 1st Conference on Lifelong Learning Agents, 2022

    2022

  24. [32]

    Tool unlearning for tool-augmented llms,

    J. Cheng and H. Amiri, “Tool unlearning for tool-augmented llms,” arXiv preprint arXiv:2502.01083, 2025

    2025 arXiv

  25. [33]

    Robust speech recognition via large-scale weak supervision,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” in International conference on machine learning , 2023

    2023

  26. [34]

    Cognivoice: Multi- modal and multilingual fusion networks for mild cognitive impair- ment assessment from spontaneous speech,

    J. Cheng, M. Elgaar, N. Vakil, and H. Amiri, “Cognivoice: Multi- modal and multilingual fusion networks for mild cognitive impair- ment assessment from spontaneous speech,” in Interspeech 2024, 2024

    2024

  27. [35]

    Mu-bench: A multitask multi- modal benchmark for machine unlearning,

    J. Cheng and H. Amiri, “Mu-bench: A multitask multi- modal benchmark for machine unlearning,” arXiv preprint arXiv:2406.14796, 2024

    2024 arXiv

  28. [36]

    Towards certified un- learning for deep neural networks,

    B. Zhang, Y . Dong, T. Wang, and J. Li, “Towards certified un- learning for deep neural networks,” in Forty-first International Conference on Machine Learning, 2024

    2024

  29. [37]

    Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation,

    C. Fan, J. Liu, Y . Zhang, E. Wong, D. Wei, and S. Liu, “Salun: Empowering machine unlearning via gradient-based weight saliency in both image classification and generation,” in International Conference on Learning Representations, 2024

    2024

  30. [38]

    Can bad teaching induce forgetting? unlearning in deep networks using an incompetent teacher,

    V . S. Chundawat, A. K. Tarun, M. Mandal, and M. Kankanhalli, “Can bad teaching induce forgetting? unlearning in deep networks using an incompetent teacher,” Proceedings of the AAAI Confer- ence on Artificial Intelligence, 2023

    2023

  31. [39]

    Superb: Speech processing universal performance benchmark,

    S. wen Yang, P.-H. Chi, Y .-S. Chuang, and Others, “Superb: Speech processing universal performance benchmark,” in Inter- speech 2021, 2021, pp. 1194–1198

    2021

  32. [40]

    Speech commands: A public dataset for single-word speech recognition

    P. Warden, “Speech commands: A public dataset for single-word speech recognition.” Dataset available online, 2017

    2017

  33. [41]

    V oxceleb: Large-scale speaker verification in the wild,

    A. Nagrani, J. S. Chung, W. Xie, and A. Zisserman, “V oxceleb: Large-scale speaker verification in the wild,”Computer Speech & Language, vol. 60, p. 101027, 2020

    2020

  34. [42]

    wav2vec 2.0: A framework for self-supervised learning of speech representa- tions,

    A. Baevski, H. Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech representa- tions,” 2020

    2020

  35. [43]

    Hubert: Self-supervised speech represen- tation learning by masked prediction of hidden units,

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdi- nov, and A. Mohamed, “Hubert: Self-supervised speech represen- tation learning by masked prediction of hidden units,” 2021

    2021

  36. [44]

    Curricu- lum learning,

    Y . Bengio, J. Louradour, R. Collobert, and J. Weston, “Curricu- lum learning,” in Proceedings of the 26th Annual International Conference on Machine Learning, 2009

    2009

  37. [45]

    Superloss: A generic loss for robust curriculum learning,

    T. Castells, P. Weinzaepfel, and J. Revaud, “Superloss: A generic loss for robust curriculum learning,” in Advances in Neural Infor- mation Processing Systems, 2020

    2020

  38. [46]

    Jogging the memory of un- learned LLMs through targeted relearning attacks,

    S. Hu, Y . Fu, S. Wu, and V . Smith, “Jogging the memory of un- learned LLMs through targeted relearning attacks,” in Interna- tional Conference on Learning Representations, 2025

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.