Pith. sign in

REVIEW 3 major objections 4 minor 62 references

UniPET-SPK: A Unified Framework for Parameter-Efficient Tuning of Pre-trained Speech Models for Robust Speaker Verification

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a gated mixture of adapter and prompt tuning outperforms full fine-tuning for speaker verification while updating only 5.4% of parameters.

desk verdict A genuinely new gated fusion of adapters and prompts that beats full fine-tuning, but missing [54] and same-trial hyperparameter selection make the 'best reported' claim provisional. read the letter →

arxiv 2501.16542 v1 pith:JGQ7UPH5 submitted 2025-01-27 eess.AS cs.LGcs.SD

Mufan Sang , John H. L. Hansen This is my paper · ORCID
classification eess.AScs.LGcs.SD
keywords speakerverificationparameter-efficienttuningadapter-tuningprompt-tuninggatingmechanismpre-trainedspeechmodelstransferlearningWavLM
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

This paper tries to show that large pre-trained speech models can be adapted to speaker verification more cheaply and more accurately by tuning only a small set of extra parameters instead of fine-tuning the whole model. It proposes three methods: an adapter that works inside and across Transformer layers, a deep prompt that prepends learnable vectors at every layer, and a unified framework, UniPET-SPK, that lets each layer learn how much to rely on each. The paper reports that UniPET-SPK beats full fine-tuning and the other PET baselines on English, Chinese, and forensic datasets while updating only 5.4% of model parameters. If true, the result matters because it would make large self-supervised speech backbones practical to reuse across many speaker-verification tasks and domains.

What carries the argument

The load-bearing object is the UniPET-SPK framework, which couples two PET submodules per Transformer layer: an Inner+Inter Adapter (a bottleneck adapter inserted in parallel with each feed-forward block, plus a small adapter after the weighted sum of all layer outputs) and Deep Speaker Prompting (trainable vectors prepended to the input of every layer). A per-layer gating function, computed from the layer's hidden states by a small feed-forward network with sigmoid activation, scales the prompt tokens and the adapter outputs before they enter the frozen Transformer. The gates are trained together with the adapters, prompts, and speaker-verification backend, so the model can dynamically choose the mixture of PET methods layer by layer and dataset by dataset. This carries the argument by turning 'which PET method?' into a learned per-layer interpolation rather than a discrete choice.

What would settle it

Fix UniPET-SPK's hyperparameters using a validation split that is disjoint from VoxCeleb1-O (for example, sweep on VoxCeleb1-E and then test on VoxCeleb1-O) and add the dual adapter-and-prompt baseline [54] to the comparison; if EER no longer beats full fine-tuning by the reported margin, the claimed advantage is at least in part an artifact of selection on the evaluation set.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that parameter-efficient tuning is not just a compression trick: a carefully mixed PET configuration can outperform full fine-tuning for speaker verification. With WavLM Base+ and a TDNN backend, UniPET-SPK reaches 1.59% EER on VoxCeleb1-O versus 1.82% for full fine-tuning, a 12.6% relative EER improvement and a 19.3% relative minDCF improvement, while updating 5.1M of the 94.7M backbone parameters (5.4%) plus the backend. The same ordering holds with HuBERT Base and becomes larger on the smaller, domain-mismatched CN-Celeb1 and 1st48-UTD corpora, where full fine-tuning overfits. The paper further argues that the gating mechanism, not merely the extra trainable parameters, drives the gains, because a non-gated combination of the same adapters and prompts performs worse.

Load-bearing premise

The headline comparison rests on whether the evaluation protocol is fair: UniPET-SPK's hyperparameters were chosen by sweeping on the same VoxCeleb1 data used for the headline VoxCeleb1-O result, so if those choices are overfit to that trial list the reported margins could shrink or disappear on new trials.

Editorial extensions

If this is right

  • Updating 5.4% of parameters can beat full fine-tuning, so the per-task storage and deployment cost of a speaker-verification system built on an SSL backbone drops by roughly an order of magnitude.
  • The improvement over full fine-tuning is largest on small, domain-shifted datasets such as CN-Celeb1 and 1st48-UTD, suggesting PET is especially useful when training data is scarce.
  • Because the pre-trained backbone stays frozen and only prompts, adapters, gates, and the backend are stored per task, one shared model can serve many speaker-verification tasks or domains.
  • Ablations in the paper show that combining adapters and prompts without gates performs worse than the gated version, so the learned per-layer mixture is what carries the benefit.
  • Deep Speaker Prompting alone, with 0.3% of the parameters, comes close to full fine-tuning on VoxCeleb1-O, so an ultra-light adaptation path exists when storage is the binding constraint.

Reading between the lines

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

  • A natural next test is whether the same gated mixture transfers to other speech tasks such as emotion recognition, language identification, or ASR; the paper's layer-weight analysis suggests the mechanism is generic, but that is an inference beyond its reported experiments.
  • Because the hyperparameters were selected on VoxCeleb1 and evaluated on the same VoxCeleb1-O trial list, an out-of-sample check on held-out trials or on a different language would clarify how much of the reported margin is real; this is an editorial caution, not a claim in the paper.
  • In deployment, the framework suggests a multi-tenant design in which one frozen backbone is shared across clients or domains, each with its own small gated adapter and prompt bundle; the paper does not discuss this use case.
  • Including the dual adapter-and-prompt method cited as [54] in the comparison tables would complete the claimed superiority over other PET methods; the paper does not compare against it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes three parameter-efficient tuning (PET) methods for adapting frozen self-supervised speech encoders to speaker verification: an Inner+Inter Adapter, a Deep Speaker Prompting method, and UniPET-SPK, which combines both through per-layer learnable gates. Experiments are conducted with WavLM Base+ and HuBERT Base backbones on VoxCeleb1, CN-Celeb1, and the 1st48-UTD forensic corpus. The central claim, stated in the abstract and Table I, is that UniPET-SPK consistently outperforms full fine-tuning, its two component methods, and other PET baselines while updating only 5.4% of the pre-trained model parameters. The headline result is 1.59% EER on VoxCeleb1-O with WavLM Base+ and a TDNN backend, compared with 1.82% EER for full fine-tuning.

Significance. If the reported results hold under a fair protocol, the paper makes a useful contribution: it provides a simple and effective unified gating mechanism, demonstrates that PET can beat full fine-tuning for speaker verification, and validates the approach across two SSL backbones, three datasets, and both linear and TDNN backends. The gating ablation (UniPET-SPK w/o gate vs. UniPET-SPK in Table I) directly supports the claimed role of the learned gates, and the cross-dataset results on CN-Celeb1 and 1st48-UTD give evidence of generalization beyond VoxCeleb. The paper is an empirical study rather than a derivation, so the main risk is not internal inconsistency but whether the comparison set and evaluation protocol fully support the 'consistently outperforms other PET methods' claim.

major comments (3)
  1. [Section II-C and Tables I, V, VI] Reference [54] is the closest prior work to UniPET-SPK, since it also combines speaker prompt tuning with adapters, yet it is cited in Section II-C but never appears in any comparison table. Because the abstract and Table I claim consistent superiority over 'other parameter-efficient tuning methods,' the comparison set is incomplete. Please add [54] to the experimental tables or explicitly justify its exclusion.
  2. [Section VII-B and Table I] The hyperparameters that define UniPET-SPK (adapter dimension 256, scaling factor 0.5, prompt length 30) are selected in Section VII-B by sweeps whose EER/minDCF are reported on VoxCeleb1, which includes the same VoxCeleb1-O trial list used for the headline result in Table I. This makes the VoxCeleb1-O number configuration-selected rather than a held-out evaluation. The VoxCeleb1-E/H margins and the other datasets partially mitigate the concern, but do not remove it. Please state explicitly which trial is used for the ablations, move hyperparameter selection to a development split, or show that neighboring hyperparameter values yield similar margins.
  3. [Tables I, V, VI and Section VII] No repeated-seed results, error bars, or significance tests are reported, and several key margins are small. For example, with HuBERT Base and TDNN on VoxCeleb1-O, UniPET-SPK achieves 1.89% EER versus 2.04% for Inner+Inter Adapter and 1.98% for full fine-tuning; with WavLM Base+ and TDNN on VoxCeleb1-O, UniPET-SPK w/o gate is 1.80% versus 1.59% with gate. Without variance estimates or multiple seeds, the gating benefit and the claim of consistent outperformance are not statistically established. Please add mean±std over repeated runs or an appropriate significance test for the main comparisons.
minor comments (4)
  1. [Section V] There are typos in the text: 'learnble' should be 'learnable' and 'Deep Speaker Promoting' should be 'Deep Speaker Prompting.'
  2. [Section IV, Eq. (7)] The notation in Eq. (7) and the surrounding text does not make clear whether the prompt tokens P_i are inserted before the next layer while the latent representations Z_i from the current layer are discarded, or how the new prompts are concatenated at each layer boundary; please define the exact sequence explicitly.
  3. [Table I] The '# Params' column mixes adapter/prompt parameters and SV backend parameters with a '+' sign, while the main text states 'updating only 5.4% of the parameters' using only the first term; please state explicitly whether the reported percentage includes the backend parameters.
  4. [Figure 1] The caption says the area of each circle is proportional to the total number of tunable parameters, but the figure is difficult to read from the plotted markers alone; consider adding numerical labels to each circle or a clearly readable parameter axis.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: UniPET-SPK hyperparameters are tuned on the same VoxCeleb1-O trial list used for the headline result; independent datasets limit the damage.

  1. fitted input called prediction [Section VI.A.1 and Section VII-B (Tables II-IV), with the headline claim in the Abstract and Table I]
    "For all ablation study experiments in Sec. VII-B, all systems are trained on the VoxCeleb1 development set... We evaluate performance of all ASV systems on three evaluation trials: VoxCeleb1-O, VoxCeleb1-E, and VoxCeleb1-H. ... As shown in Table II, Inner+Inter Adapter achieves the best performance when the dimension size is 256. ... Table III shows that the parallel adapter achieves the best performance with a fixed scale at 0.5. ... the Deep Speaker Prompting obtains the best performance when the length is 30. ... Therefore, we choose a prompt length of 30 in all experiments."

    The three UniPET-SPK configuration choices (adapter dimension 256, scaling 0.5, prompt length 30) are selected by sweeping EER on VoxCeleb1, which includes the VoxCeleb1-O trial list; the Abstract and Table I then advertise the VoxCeleb1-O EER of 1.59% versus 1.82% for full fine-tuning as evidence of consistent superiority. The reported VoxCeleb1-O margin is therefore not an independent prediction for that trial list: the configuration was chosen using that same trial list as the selection objective. The result is not definitionally forced because the Table I model is trained on VoxCeleb2, and the VoxCeleb1-E/H rows plus CN-Celeb and 1st48-UTD evaluations provide separate evidence, but the headline O-number is partly a selection artifact rather than a fresh test.

full rationale

This paper is an empirical systems comparison, not a mathematical derivation, so there is no definitional circularity of the form 'Eq. X = Eq. Y by construction.' The authors' citation of their own previous adapter work [29] is contextual and not load-bearing; the Inner+Inter Adapter, Deep Speaker Prompting, and gating mechanism are specified by the paper's own equations (1)-(10). The closest dual-PET baseline [54] is cited but absent from the comparison tables, which is a completeness or fairness concern rather than a circularity concern. The one genuine circularity-adjacent issue is that the UniPET-SPK hyperparameters are selected in Section VII-B on VoxCeleb1, and the headline 'consistently outperforms' result on VoxCeleb1-O in Table I is evaluated on the same VoxCeleb1-O trial list. This makes the O-number partially a product of selection on the test metric. The damage is mitigated by the VoxCeleb1-E and VoxCeleb1-H rows, which were not the selection target, and by the fully independent CN-Celeb and 1st48-UTD evaluations, which use the same configuration but different evaluation data and still show consistent gains. The absence of repeated-seed variance reporting is an experimental rigor issue, not circularity. On balance, the central claim retains independent content, so the score is 4 rather than higher.

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

The empirical contribution depends on a handful of hyperparameters selected by validation on the evaluation benchmarks, most importantly the adapter dimension, the fixed adapter scaling factor, and the prompt length. The approach also assumes the standard SSL transfer-learning setup (frozen encoder, weighted-sum pooling) and that the reimplemented baselines were tuned fairly. No new physical or mathematical entities are introduced.

free parameters (4)
  • Adapter bottleneck dimension = 256
    Selected by sweep on VoxCeleb1 (Table II); dimension 256 gave the best EER/minDCF on VoxCeleb1-O.
  • Adapter scaling factor s = 0.5
    Selected by sweep on VoxCeleb1 (Table III); fixed scale 0.5 gives the best performance over learnable and other fixed scales.
  • Speaker prompt length = 30
    Selected by sweep on VoxCeleb1 (Table IV); 30 prompt tokens gave the best EER/minDCF across the tested range.
  • LoRA rank for baseline = 64
    For the LoRA baseline, rank r=64 'achieves the best performance' per Section VI-B.
assumptions (4)
  • domain assumption Pre-trained SSL speech models (WavLM Base+, HuBERT Base) provide frozen feature extractors whose representations are suitable for speaker verification.
    The entire PET approach presupposes that frozen SSL representations are useful and need only light adaptation; this is the standard SUPERB-style assumption.
  • domain assumption The evaluation protocols (VoxCeleb1-O/E/H, CN-Celeb1, 1st48-UTD) are reliable measures of speaker verification performance.
    Standard benchmarks are used, but 1st48-UTD is private and small, and the paper does not report variance over trials.
  • domain assumption The reimplemented baselines (Houlsby, E+L, LoRA) are tuned fairly and equivalently to the proposed methods.
    Authors state the same training configurations are used, but no code or detailed baseline hyperparameters are provided.
  • domain assumption Freezing the CNN encoder during full fine-tuning is a valid configuration.
    Full fine-tuning keeps the convolutional encoder fixed following SUPERB; this limits the capacity of fine-tuning and may favor PET methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UniPET-SPK: A Unified Framework for Parameter-Efficient Tuning of Pre-trained Speech Models for Robust Speaker Verification." pith.science (2026). https://pith.science/paper/JGQ7UPH5

@misc{pith2026250116542,
  author       = {Pith},
  title        = {Pith review of: UniPET-SPK: A Unified Framework for Parameter-Efficient Tuning of Pre-trained Speech Models for Robust Speaker Verification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JGQ7UPH5}},
  note         = {Machine review of arXiv:2501.16542}
}
read the original abstract

With excellent generalization ability, SSL speech models have shown impressive performance on various downstream tasks in the pre-training and fine-tuning paradigm. However, as the size of pre-trained models grows, fine-tuning becomes practically unfeasible due to expanding computation and storage requirements and the risk of overfitting. This study explores parameter-efficient tuning (PET) methods for adapting large-scale pre-trained SSL speech models to speaker verification task. Correspondingly, we propose three PET methods: (i)an adapter-tuning method, (ii)a prompt-tuning method, and (iii)a unified framework that effectively incorporates adapter-tuning and prompt-tuning with a dynamically learnable gating mechanism. First, we propose the Inner+Inter Adapter framework, which inserts two types of adapters into pre-trained models, allowing for adaptation of latent features within the intermediate Transformer layers and output embeddings from all Transformer layers, through a parallel adapter design. Second, we propose the Deep Speaker Prompting method that concatenates trainable prompt tokens into the input space of pre-trained models to guide adaptation. Lastly, we propose the UniPET-SPK, a unified framework that effectively incorporates these two alternate PET methods into a single framework with a dynamic trainable gating mechanism. The proposed UniPET-SPK learns to find the optimal mixture of PET methods to match different datasets and scenarios. We conduct a comprehensive set of experiments on several datasets to validate the effectiveness of the proposed PET methods. Experimental results on VoxCeleb, CN-Celeb, and 1st 48-UTD forensic datasets demonstrate that the proposed UniPET-SPK consistently outperforms the two PET methods, fine-tuning, and other parameter-efficient tuning methods, achieving superior performance while updating only 5.4% of the parameters.

Figures

Figures reproduced from arXiv: 2501.16542 by the authors.

Figure 1
Figure 1. Performance of parameter-efficient tuning approaches on Vox1-O. The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the parameter-efficient tuning methods. (a) Adapter, which inserts learnable lightweight networks into a pre-trained model; (b) Prompt, [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the pre-trained model with Inner+Inter Adapter framework, and the detailed architectures of sequential and parallel adapters. During [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Overview of the UniPET-SPK framework. During fine-tuning, the pre-trained model is frozen, only the prepended speaker prompts, Inner-layer [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Layer weight analysis on the VoxCeleb dataset. Layers from 1 to 12 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 52 canonical work pages

  1. [54]

    Dual parameter-efficient fine-tuning for speaker representation via speaker prompt tuning and adapters,

    Z. Li, M.-W. Mak, and H. M.-L. Meng, “Dual parameter-efficient fine-tuning for speaker representation via speaker prompt tuning and adapters,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2024, pp. 10 751–10 755

    work page 2024

  2. [1]

    X-vectors: Robust dnn embeddings for speaker recognition,

    D. Snyder, D. Garcia-Romero, G. Sell, D. Povey, and S. Khudanpur, “X-vectors: Robust dnn embeddings for speaker recognition,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. , 2018, pp. 5329–5333

    work page 2018

  3. [2]

    But system description to voxceleb speaker recognition challenge 2019,

    H. Zeinali, S. Wang, A. Silnova, P. Mat ˇejka, and O. Plchot, “But system description to voxceleb speaker recognition challenge 2019,” arXiv preprint arXiv:1910.12592 , 2019

    arXiv 2019

  4. [3]

    ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,

    B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA-TDNN: Emphasized Channel Attention, Propagation and Aggregation in TDNN Based Speaker Verification,” inProc. ISCA Interspeech, 2020, pp. 3830– 3834

    work page 2020

  5. [4]

    Deep speaker embedding extraction with channel-wise feature responses and additive supervision softmax loss function

    J. Zhou, T. Jiang, Z. Li, L. Li, and Q. Hong, “Deep speaker embedding extraction with channel-wise feature responses and additive supervision softmax loss function.” in Proc. ISCA Interspeech, 2019, pp. 2883–2887

    work page 2019

  6. [5]

    Frequency and temporal convolutional attention for text-independent speaker recognition,

    S. Yadav and A. Rai, “Frequency and temporal convolutional attention for text-independent speaker recognition,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. , 2020, pp. 6794–6798

    work page 2020

  7. [6]

    Multi-Frequency Information Enhanced Channel Attention Module for Speaker Representation Learning,

    M. Sang and J. H. L. Hansen, “Multi-Frequency Information Enhanced Channel Attention Module for Speaker Representation Learning,” in Proc. ISCA Interspeech , 2022, pp. 321–325

    work page 2022

  8. [7]

    Simple attention module based speaker verification with iterative noisy label detection,

    X. Qin, N. Li, C. Weng, D. Su, and M. Li, “Simple attention module based speaker verification with iterative noisy label detection,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process , 2022, pp. 6722–6726

    work page 2022

Show all 62 references
  1. [8]

    S-vectors and tesa: Speaker embeddings and a speaker authenticator based on transformer encoder,

    N. J. M. S. Mary, S. Umesh, and S. V . Katta, “S-vectors and tesa: Speaker embeddings and a speaker authenticator based on transformer encoder,” IEEE Trans. Audio, Speech, Lang. Process. , vol. 30, pp. 404– 413, 2021

    2021

  2. [9]

    MFA-Conformer: Multi-scale Feature Aggregation Conformer for Au- tomatic Speaker Verification,

    Y . Zhang, Z. Lv, H. Wu, S. Zhang, P. Hu, Z. Wu, H. yi Lee, and H. Meng, “MFA-Conformer: Multi-scale Feature Aggregation Conformer for Au- tomatic Speaker Verification,” inProc. ISCA Interspeech, 2022, pp. 306– 310

    2022

  3. [10]

    Improving transformer-based networks with locality for automatic speaker verifica- tion,

    M. Sang, Y . Zhao, G. Liu, J. H. L. Hansen, and J. Wu, “Improving transformer-based networks with locality for automatic speaker verifica- tion,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. IEEE, 2023, pp. 1–5

    2023

  4. [11]

    Unispeech-sat: Universal speech representation learning with speaker aware pre-training,

    S. Chen, Y . Wu, C. Wang, Z. Chen, Z. Chen, S. Liu, J. Wu, Y . Qian, F. Wei, J. Li et al. , “Unispeech-sat: Universal speech representation learning with speaker aware pre-training,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2022, pp. 6152–6156

    2022

  5. [12]

    Contrastive self-supervised learning for text-independent speaker verification,

    H. Zhang, Y . Zou, and H. Wang, “Contrastive self-supervised learning for text-independent speaker verification,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2021, pp. 6713–6717

    2021

  6. [13]

    Self-supervised speaker verification with simple siamese network and self-supervised regularization,

    M. Sang, H. Li, F. Liu, A. O. Arnold, and L. Wan, “Self-supervised speaker verification with simple siamese network and self-supervised regularization,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process., 2022, pp. 6127–6131

    2022

  7. [14]

    Generative adversarial speaker embedding networks for domain robust end-to-end speaker verification,

    G. Bhattacharya, J. Monteiro, J. Alam, and P. Kenny, “Generative adversarial speaker embedding networks for domain robust end-to-end speaker verification,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process., 2019, pp. 6226–6230

    2019

  8. [15]

    Cross-domain adaptation with discrepancy minimization for text-independent forensic speaker verification,

    Z. Wang, W. Xia, and J. H. L. Hansen, “Cross-domain adaptation with discrepancy minimization for text-independent forensic speaker verification,” in Proc. ISCA Interspeech , 2020, pp. 2257–2261

    2020

  9. [16]

    Deaan: Disentangled embedding and adversarial adaptation network for robust speaker representation learning,

    M. Sang, W. Xia, and J. H. L. Hansen, “Deaan: Disentangled embedding and adversarial adaptation network for robust speaker representation learning,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2021, pp. 6169–6173

    2021

  10. [17]

    Speaker recognition by machines and humans: A tutorial review,

    J. H. L. Hansen and T. Hasan, “Speaker recognition by machines and humans: A tutorial review,” IEEE Signal Process. Mag. , vol. 32, no. 6, pp. 74–99, 2015

    2015

  11. [18]

    Speaker verification with short utterances: a review of challenges, trends and opportunities,

    A. Poddar, M. Sahidullah, and G. Saha, “Speaker verification with short utterances: a review of challenges, trends and opportunities,” IET Biometrics, vol. 7, no. 2, pp. 91–101, 2018

    2018

  12. [19]

    Open-Set Short Utterance Forensic Speaker Verification Using Teacher-Student Network with Explicit Inductive Bias,

    M. Sang, W. Xia, and J. H. L. Hansen, “Open-Set Short Utterance Forensic Speaker Verification Using Teacher-Student Network with Explicit Inductive Bias,” in Proc. ISCA Interspeech , 2020, pp. 2262– 2266

    2020

  13. [20]

    Multi-source domain adaptation for text- independent forensic speaker recognition,

    Z. Wang and J. H. Hansen, “Multi-source domain adaptation for text- independent forensic speaker recognition,” IEEE Trans. Audio, Speech, Lang. Process., vol. 30, pp. 60–75, 2021

    2021

  14. [21]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst. , vol. 30, 2017

    2017

  15. [22]

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

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A framework for self-supervised learning of speech representations,” Proc. Int. Conf. Neural Inf. Process. Syst , vol. 33, pp. 12 449–12 460, 2020

    2020

  16. [23]

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

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “Hubert: Self-supervised speech representation learning by masked prediction of hidden units,”IEEE Trans. Audio, Speech, Lang. Process., vol. 29, pp. 3451–3460, 2021

    2021

  17. [24]

    Wavlm: Large-scale self-supervised pre- training for full stack speech processing,

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao et al. , “Wavlm: Large-scale self-supervised pre- training for full stack speech processing,” IEEE Journal of Selected Topics in Signal Processing , vol. 16, no. 6, pp. 1505–1518, 2022

    2022

  18. [25]

    Parameter-efficient transfer learning for nlp,

    N. Houlsby, A. Giurgiu, S. Jastrzebski, B. Morrone, Q. De Laroussilhe, A. Gesmundo, M. Attariyan, and S. Gelly, “Parameter-efficient transfer learning for nlp,” in Proc. Int. Conf. Mach. Learn., 2019, pp. 2790–2799

    2019

  19. [26]

    Efficient adapter transfer of self- supervised speech models for automatic speech recognition,

    B. Thomas, S. Kessler, and S. Karout, “Efficient adapter transfer of self- supervised speech models for automatic speech recognition,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2022, pp. 7102–7106

    2022

  20. [27]

    DRAFT: A Novel Framework to Reduce Domain Shifting in Self-supervised Learning and Its Application to Children’s ASR,

    R. Fan and A. Alwan, “DRAFT: A Novel Framework to Reduce Domain Shifting in Self-supervised Learning and Its Application to Children’s ASR,” in Proc. ISCA Interspeech , 2022, pp. 4900–4904

    2022

  21. [28]

    Chapter: Exploiting convolu- tional neural network adapters for self-supervised speech models,

    Z.-C. Chen, Y .-S. Sung, and H.-y. Lee, “Chapter: Exploiting convolu- tional neural network adapters for self-supervised speech models,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process Workshops , 2023, pp. 1–5

    2023

  22. [29]

    Efficient adapter tuning of pre-trained speech models for automatic speaker verification,

    M. Sang and J. H. L. Hansen, “Efficient adapter tuning of pre-trained speech models for automatic speaker verification,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2024, pp. 12 131–12 135

    2024

  23. [30]

    Pre- train, prompt, and predict: A systematic survey of prompting methods in natural language processing,

    P. Liu, W. Yuan, J. Fu, Z. Jiang, H. Hayashi, and G. Neubig, “Pre- train, prompt, and predict: A systematic survey of prompting methods in natural language processing,” ACM Comput. Surv., vol. 55, no. 9, jan 2023

    2023

  24. [31]

    Prefix-tuning: Optimizing continuous prompts for generation,

    X. L. Li and P. Liang, “Prefix-tuning: Optimizing continuous prompts for generation,” in ACL/IJCNLP (1). Association for Computational Linguistics, 2021, pp. 4582–4597

    2021

  25. [32]

    The power of scale for parameter-efficient prompt tuning,

    B. Lester, R. Al-Rfou, and N. Constant, “The power of scale for parameter-efficient prompt tuning,” in Proc. Conf. on Empir. Methods Natural Lang. Process., 2021, pp. 3045–3059

    2021

  26. [33]

    Unipelt: A unified framework for parameter-efficient language model tuning,

    Y . Mao, L. Mathias, R. Hou, A. Almahairi, H. Ma, J. Han, W. T. Yih, and M. Khabsa, “Unipelt: A unified framework for parameter-efficient language model tuning,” in Proc. 60th Annu. Meeting Assoc. Comput. Linguistics., 2022, pp. 6253–6264

    2022

  27. [34]

    Parameter-efficient transfer learning of pre-trained trans- former models for speaker verification using adapters,

    J. Peng, T. Stafylakis, R. Gu, O. Plchot, L. Mo ˇsner, L. Burget, and J. ˇCernock`y, “Parameter-efficient transfer learning of pre-trained trans- former models for speaker verification using adapters,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2023, pp. 1–5

    2023

  28. [35]

    wav2vec: Unsupervised Pre-Training for Speech Recognition,

    S. Schneider, A. Baevski, R. Collobert, and M. Auli, “wav2vec: Unsupervised Pre-Training for Speech Recognition,” in Proc. ISCA Interspeech, 2019, pp. 3465–3469

    2019

  29. [36]

    Towards a unified view of parameter-efficient transfer learning,

    J. He, C. Zhou, X. Ma, T. Berg-Kirkpatrick, and G. Neubig, “Towards a unified view of parameter-efficient transfer learning,” International Conference on Learning Representations , 2022

    2022

  30. [37]

    Vl-adapter: Parameter-efficient transfer learning for vision-and-language tasks,

    Y .-L. Sung, J. Cho, and M. Bansal, “Vl-adapter: Parameter-efficient transfer learning for vision-and-language tasks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2022, pp. 5227–5237

    2022

  31. [38]

    Adapt- former: Adapting vision transformers for scalable visual recognition,

    S. Chen, C. Ge, Z. Tong, J. Wang, Y . Song, J. Wang, and P. Luo, “Adapt- former: Adapting vision transformers for scalable visual recognition,” Proc. Adv. Neural Inf. Process. Syst. , vol. 35, pp. 16 664–16 678, 2022. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    2022

  32. [39]

    Large-Scale Multilingual Speech Recognition with a Streaming End-to-End Model,

    A. Kannan, A. Datta, T. N. Sainath, E. Weinstein, B. Ramabhadran, Y . Wu, A. Bapna, Z. Chen, and S. Lee, “Large-Scale Multilingual Speech Recognition with a Streaming End-to-End Model,” inProc. ISCA Interspeech, 2019, pp. 2130–2134

    2019

  33. [40]

    Adapt-and-Adjust: Over- coming the Long-Tail Problem of Multilingual Speech Recognition,

    G. I. Winata, G. Wang, C. Xiong, and S. Hoi, “Adapt-and-Adjust: Over- coming the Long-Tail Problem of Multilingual Speech Recognition,” in Proc. ISCA Interspeech , 2021, pp. 2451–2455

    2021

  34. [41]

    Exploiting adapters for cross-lingual low-resource speech recognition,

    W. Hou, H. Zhu, Y . Wang, J. Wang, T. Qin, R. Xu, and T. Shinozaki, “Exploiting adapters for cross-lingual low-resource speech recognition,” IEEE Trans. Audio, Speech, Lang. Process., vol. 30, pp. 317–329, 2021

    2021

  35. [42]

    Parameter efficient transfer learning for various speech processing tasks,

    S. Otake, R. Kawakami, and N. Inoue, “Parameter efficient transfer learning for various speech processing tasks,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2023, pp. 1–5

    2023

  36. [43]

    Language models are unsupervised multitask learners,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskever et al., “Language models are unsupervised multitask learners,” OpenAI blog , vol. 1, no. 8, p. 9, 2019

    2019

  37. [44]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert- V oss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S. Gray, B. C...

    2020

  38. [45]

    Gpt understands, too,

    X. Liu, Y . Zheng, Z. Du, M. Ding, Y . Qian, Z. Yang, and J. Tang, “Gpt understands, too,” arXiv preprint arXiv:2103.10385 , 2021

    2021 arXiv

  39. [46]

    P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks,

    X. Liu, K. Ji, Y . Fu, W. Tam, Z. Du, Z. Yang, and J. Tang, “P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks,” in Proc. 60th Annu. Meeting Assoc. Comput. Linguistics. , 2022, pp. 61– 68

    2022

  40. [47]

    Visual prompt tuning,

    M. Jia, L. Tang, B.-C. Chen, C. Cardie, S. Belongie, B. Hariharan, and S.-N. Lim, “Visual prompt tuning,” in Proc. Eur. Conf. Comput. Vis. Springer, 2022, pp. 709–727

    2022

  41. [48]

    Learning transferable visual models from natural language supervi- sion,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learning transferable visual models from natural language supervi- sion,” in Proc. Int. Conf. Mach. Learn. , vol. 139, 18–24 Jul 2021, p...

    2021

  42. [49]

    Learning to prompt for vision- language models,

    K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to prompt for vision- language models,” International Journal of Computer Vision , vol. 130, no. 9, pp. 2337–2348, 2022

    2022

  43. [50]

    An Exploration of Prompt Tuning on Generative Spoken Language Model for Speech Processing Tasks,

    K.-W. Chang, W.-C. Tseng, S.-W. Li, and H. yi Lee, “An Exploration of Prompt Tuning on Generative Spoken Language Model for Speech Processing Tasks,” in Proc. ISCA Interspeech , 2022, pp. 5005–5009

    2022

  44. [51]

    Speechprompt v2: Prompt tuning for speech classification tasks,

    K.-W. Chang, Y .-K. Wang, H. Shen, I.-t. Kang, W.-C. Tseng, S.- W. Li, and H.-y. Lee, “Speechprompt v2: Prompt tuning for speech classification tasks,” arXiv preprint arXiv:2303.00733 , 2023

    2023 arXiv

  45. [52]

    Integrated parameter-efficient tuning for general-purpose audio models,

    J.-h. Kim, J. Heo, H.-s. Shin, C.-y. Lim, and H.-J. Yu, “Integrated parameter-efficient tuning for general-purpose audio models,” arXiv preprint arXiv:2211.02227, 2022

    2022 arXiv

  46. [53]

    Ast: Audio spectrogram trans- former,

    Y . Gong, Y .-A. Chung, and J. Glass, “Ast: Audio spectrogram trans- former,” in Proc. ISCA Interspeech , 2021, pp. 571–575

    2021

  47. [55]

    SUPERB: Speech Processing Universal PERformance Benchmark,

    S. wen Yang, P.-H. Chi, Y .-S. Chuang, C.-I. J. Lai, K. Lakhotia, Y . Y . Lin, A. T. Liu, J. Shi, X. Chang, G.-T. Lin, T.-H. Huang, W.-C. Tseng, K. tik Lee, D.-R. Liu, Z. Huang, S. Dong, S.-W. Li, S. Watanabe, A. Mohamed, and H. yi Lee, “SUPERB: Speech Processing Universal PER...

    2021

  48. [56]

    V oxceleb: A large-scale speaker identification dataset,

    A. Nagrani, J. S. Chung, and A. Zisserman, “V oxceleb: A large-scale speaker identification dataset,” Proc. ISCA Interspeech, pp. 2616–2620, 2017

    2017

  49. [57]

    V oxceleb2: Deep speaker recognition,

    J. S. Chung, A. Nagrani, and A. Zisserman, “V oxceleb2: Deep speaker recognition,” Proc. ISCA Interspeech , pp. 1086–1090, 2018

    2018

  50. [58]

    Cn-celeb: a challenging chinese speaker recogni- tion dataset,

    Y . Fan, J. Kang, L. Li, K. Li, H. Chen, S. Cheng, P. Zhang, Z. Zhou, Y . Cai, and D. Wang, “Cn-celeb: a challenging chinese speaker recogni- tion dataset,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Process. , 2020, pp. 7604–7608

    2020

  51. [59]

    Understanding the difficulty of training deep feedforward neural networks,

    X. Glorot and Y . Bengio, “Understanding the difficulty of training deep feedforward neural networks,” in Proceedings of the thirteenth international conference on artificial intelligence and statistics . JMLR Workshop and Conference Proceedings, 2010, pp. 249–256

    2010

  52. [60]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  53. [61]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 4690–4699

    2019

  54. [62]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” in International Conference on Learning Representations , 2022

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.