Pith. sign in

REVIEW 5 major objections 4 minor 33 references

Frame-level Temporal Difference Learning for Partial Deepfake Speech Detection

T0 review · 5 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Partially faked speech can be detected from erratic frame-direction changes in acoustic embeddings, using only utterance-level labels; the reported equal error rates are 0.59% on PartialSpoof and 0.03% on Half-Truth.

desk verdict A worthwhile weakly-supervised partial deepfake detection idea with solid ablations, but the headline generalization claim outruns the evidence because the short-segment regime is never directly tested. read the letter →

arxiv 2507.15101 v1 pith:7S5RRNZS submitted 2025-07-20 cs.SD cs.CReess.AS

Menglu Li , Xiao-Ping Zhang , Lian Zhao This is my paper · ORCID
classification cs.SDcs.CReess.AS
keywords partialdeepfakespeechdetectiontemporaldifferencelearningwav2vec2-XLSRweaklysupervisedadaptiveaveragepoolinganti-spoofingattentioncross-lingualgeneralization
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

Partially faked speech—utterances where only some segments are synthetic—can be detected without expensive frame-level labels. The paper shows that in wav2vec2-XLSR audio embeddings, bonafide speech moves with smoother, more consistent frame-to-frame direction, while deepfake segments fluctuate erratically. It introduces a Temporal Difference Attention Module that learns to weight frames by these irregularities at fine and coarse time scales, then averages frame predictions into an utterance verdict. On the PartialSpoof and Half-Truth datasets the model reaches equal error rates of 0.59% and 0.03%, respectively, outperforming systems that rely on transition boundaries or segment labels.

What carries the argument

The load-bearing component is the Temporal Difference Attention Module (TDAM), which operates on a refined embedding $E$ of $T'$ frames. It first computes the directional difference map $M[t] = E_{\text{conv}}[t+1] - E[t]$ for $t = 1,\dots,T'-1$ with a zero vector appended to keep the shape. This map feeds two convolutional levels: $X_1$ catches fine-grained local inconsistencies, and $X_2$ is formed by downsampling, convolving, and upsampling to capture longer-term motion. The sum $X_1 + X_2$ passes through a sigmoid to yield frame confidence scores, and the output is $Y = \sigma[l*(X_1+X_2)] \odot E$, where $l$ is a $1\times1$ convolution. A softmax over $Y$ gives frame predictions that are averaged into the utterance verdict. Adaptive average pooling first partitions the variable-length frames into $T'=4$ second segments so the pipeline accepts any input duration.

What would settle it

Sort PartialSpoof evaluation utterances into bins by the total duration of injected deepfake segments and recompute equal error rate per bin: if the shortest-duration bin shows a large EER increase, the claim that pooling preserves short-segment directional cues fails.

Watch

Extended reading notes

Core claim

The central claim is that frame-level temporal difference direction is a sufficient cue for partial deepfake detection when computed on wav2vec2-XLSR embeddings. In the paper's motivating analysis, bonafide utterances yield higher mean and lower standard deviation of the cosine similarity between adjacent normalized direction vectors, indicating natural prosodic flow, whereas deepfakes show oscillatory directional changes. On this basis, TDAM-AvgPool captures both fine-grained and long-range temporal irregularities and, trained only on utterance-level labels, reports equal error rates of 0.19% on the PartialSpoof development set, 0.59% on its evaluation set, and 0.03% on the Half-Truth test set, plus 99.99% AUC on HAD. The ablations attribute most of the gain to the TDAM module itself and to preserving directional rather than absolute differences.

Load-bearing premise

The load-bearing premise is that first-order directional irregularity in wav2vec2-XLSR frame embeddings is a generalizable marker of synthetic speech, and that adaptive average pooling to 4-second segments preserves this marker even when the injected fake segments are short.

Editorial extensions

If this is right

  • Frame-level annotations are no longer needed to train a partial deepfake detector, removing a major scalability bottleneck.
  • Smoothing of transition artifacts no longer breaks detection, because the model listens for irregularities inside manipulated regions rather than at boundaries.
  • The same detector works on fully fake utterances, since it learns artifact patterns rather than boundary patterns.
  • English and Chinese partial deepfakes are handled by one model, pointing toward cross-lingual deployment without dataset-specific tuning.
  • Ablations show that removing TDAM raises the PartialSpoof evaluation EER from 0.59% to 0.87%, so the temporal-difference cue carries the performance, not the SSL front-end alone.

Reading between the lines

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

  • Because the temporal-difference cue is defined on any frame embedding sequence, the same TDAM pipeline could be tested on other self-supervised encoders and on streaming speech where frame differences are computed causally.
  • The boxplot separation in Figs. 1 and 2 is the paper's motivating evidence but is not quantified; a natural extension is to report effect sizes or train a simple classifier on just the mean and standard deviation statistics to isolate how much of the EER comes from the statistics versus the learned attention.
  • If the directional-smoothness cue degrades for very short injected segments, duration-stratified EER would expose a practical lower bound on manipulation length the method can catch, a limit the current evaluation does not address.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 4 minor

Summary. The paper proposes a weakly supervised method for partial deepfake speech detection. It uses wav2vec2-XLSR frame embeddings, computes frame-to-frame difference features, and introduces a Temporal Difference Attention Module (TDAM) with a dual-level hierarchical representation to capture temporal irregularities without frame-level labels. Variable-length inputs are handled by adaptive average pooling. The authors report state-of-the-art results on PartialSpoof (EER 0.59%) and Half-Truth (HAD, EER 0.03%) and provide ablations showing that TDAM, the dual-level structure, directional information, and average pooling all contribute.

Significance. If the results hold, the paper offers a useful advance: it removes the need for frame-level supervision while reporting strong accuracy on partial deepfake benchmarks, and it introduces a temporal-difference perspective that is well motivated by an analysis of wav2vec2-XLSR embeddings. The public code and the internally consistent ablation study are strengths. However, the headline claim of state-of-the-art performance is currently supported only by single-run EERs with very small margins over strong baselines, and the connection between the motivating normalized-direction statistics and the actual raw-difference module is not quantitatively established. The significance is therefore conditional on additional statistical and generalization evidence, especially for short fake segments.

major comments (5)
  1. [Section VI-A, Tables I and II] All reported EERs are single-run values with no error bars, seeds, or statistical significance tests. This is load-bearing because the claimed advantage over the strongest baseline is very small on PartialSpoof (0.59% vs. 0.64% for Zhang et al. in Table I) and 0.03% vs. 0.06% on HAD. The abstract and Section VI-A say the method 'significantly outperforms' existing methods, but the paper provides no evidence of significance. Please report mean and standard deviation over multiple runs (or at least fixed seeds) and, where appropriate, a paired test.
  2. [Section III and Eq. (4)] The motivating analysis in Section III uses normalized direction vectors and cosine similarities (Eqs. (1)-(2)), but the difference map actually implemented in Eq. (4) is a raw, unnormalized difference Econv[t+1]-E[t] after a temporal convolution. No experiment demonstrates that the normalized-direction statistic separates partial fake frames from surrounding bonafide frames, nor that the model's performance is attributable to that statistic rather than to magnitude-based differences. Please quantify the motivating statistic on the PartialSpoof training/evaluation data (e.g., effect sizes or distributions) and clarify how Eq. (4) realizes the directional cue.
  3. [Section III and Section VI] The temporal-difference cue is validated only on ASVspoof2019-LA, which consists of fully fake utterances, yet the paper's central claim concerns partial deepfakes with short embedded fake segments. No duration-stratified results are reported for the PartialSpoof evaluation set. Because the evaluation set contains unseen attacks and short injected segments, the reader cannot tell whether the cue survives the pooling operation when the fake region is brief. Please report EER broken down by fake-segment duration (or by proportion of fake frames), and show corresponding frame-level localization quality, to support the claimed generalization.
  4. [Eq. (3) and Section V-B] The notation for the pooling target is inconsistent: Eq. (3) defines T' as an integer number of pooled segments, while Section V-B says 'T' is set to 4 seconds.' Since wav2vec2-XLSR embeddings are extracted every 20 ms, T'=4 seconds is not a well-defined number of segments. The effective pooled segment duration determines whether a short fake segment is preserved or averaged away, so this is not merely a wording issue. Please define T' precisely, specify the resulting segment duration, and report sensitivity of the PS evaluation EER to T'.
  5. [Table IV and Section VI-C] The text states that 'our model outperforms state-of-the-art models' in the cross-dataset study, but in the 'Training on LA' condition the proposed method has a higher EER on LA eval (1.22%) than Zhang et al. (0.83%). The claim is therefore not true as written. Please revise the claim and discuss what this weaker cross-dataset result implies for the 'strong generalizability' conclusion, especially given that the model is worse on fully fake utterances than a strong baseline when trained on LA.
minor comments (4)
  1. [Section IV-B, Eq. (5)] The symbol l in Eq. (5) is used for a 2D convolution but visually resembles the number 1 and the index variable; please use a different notation, such as W or K, to avoid confusion.
  2. [Section IV-B] The description of X1 and X2 after the 2D convolutions is incomplete: the kernel sizes, padding, strides, and how the output shape R^{C x T' x D} is obtained are not specified. Adding these details would improve reproducibility, even though the code is available.
  3. [Figure 2] The boxplots in Figure 2 are described only qualitatively. Please report the number of utterances, the means, standard deviations, and an effect-size measure so that the claimed separation between bonafide and deepfake cosine-similarity distributions is supported by numbers rather than visual inspection.
  4. [Section V-B] The sentence 'Based on duration statistics of the training datasets, we set T' to 4 seconds, matching the average utterance length' is imprecise because Eq. (3) treats T' as a count of segments. Please rephrase to state the actual number of segments used.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is benchmarked on external held-out datasets and the temporal-difference cue is empirical motivation, not a fitted prediction.

full rationale

The paper's derivation chain is self-contained and benchmark-driven. The temporal-difference hypothesis in Section III is an empirical observation computed on the ASVspoof2019-LA training set; it is not itself the fitted object, and the model's only fitted quantities are network weights plus standard hyperparameters (T', CE loss weights, learning rate) selected on training/validation data, which is ordinary model selection rather than fitting a test-set prediction. The headline EERs (0.59% on PartialSpoof, 0.03% on HAD) are measured on held-out external datasets and compared against independently published baselines, so they are not forced by construction. The self-citations (Refs. [2], [5], [7], [33]) appear only as related work and one baseline entry; no load-bearing assumption or uniqueness theorem is imported from them. The unquantified boxplots and the mismatch between the normalized directions in Eqs. (1)-(2) and the raw difference in Eq. (4) are evidentiary or consistency concerns, not circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on one empirical premise about SSL embedding dynamics, one pooling premise that short manipulations survive averaging, and one weakly supervised learning premise. No new physical entities are introduced; TDAM is a neural module, not a new postulated entity with falsifiable handles outside the paper.

free parameters (2)
  • Target pooled length T' = 4 seconds (or a frame count matching average duration; exact unit is ambiguous)
    Chosen from training-set duration statistics. It directly controls how much temporal detail survives adaptive average pooling and affects the EER results.
  • CE loss weight ratio = 9:1 (bonafide:deepfake)
    Hand-chosen to offset class imbalance in PartialSpoof and HAD. It affects the learned decision boundary and the reported performance.
assumptions (4)
  • domain assumption Frame-level wav2vec2-XLSR embeddings encode deepfake-specific first-order directional irregularity that survives fine-tuning and pooling.
    Section III, Eqs. (1)-(2), and Section IV-A: the entire method depends on this signal being present in SSL feature dynamics at 20 ms resolution; Fig. 2 shows overlapping distributions and no effect sizes are reported.
  • domain assumption Adaptive average pooling to T' segments preserves short deepfake segments.
    Section IV-A, Eq. (3): averaging frames within segments can dilute brief injected manipulations; no duration-stratified evaluation is provided.
  • domain assumption Utterance-level labels suffice to learn useful frame-level attention without localization supervision.
    Section V-B: CE loss is applied to the time-averaged prediction; the model assumes the attention map trained this way localizes artifacts rather than exploiting dataset-level confounds.
  • standard math Euclidean norm and cosine similarity are well-defined and meaningful on the SSL embedding space.
    Used in Eqs. (1)-(2) for the motivating analysis; standard background, not load-bearing beyond the empirical premise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Frame-level Temporal Difference Learning for Partial Deepfake Speech Detection." pith.science (2026). https://pith.science/paper/7S5RRNZS

@misc{pith2026250715101,
  author       = {Pith},
  title        = {Pith review of: Frame-level Temporal Difference Learning for Partial Deepfake Speech Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7S5RRNZS}},
  note         = {Machine review of arXiv:2507.15101}
}
read the original abstract

Detecting partial deepfake speech is essential due to its potential for subtle misinformation. However, existing methods depend on costly frame-level annotations during training, limiting real-world scalability. Also, they focus on detecting transition artifacts between bonafide and deepfake segments. As deepfake generation techniques increasingly smooth these transitions, detection has become more challenging. To address this, our work introduces a new perspective by analyzing frame-level temporal differences and reveals that deepfake speech exhibits erratic directional changes and unnatural local transitions compared to bonafide speech. Based on this finding, we propose a Temporal Difference Attention Module (TDAM) that redefines partial deepfake detection as identifying unnatural temporal variations, without relying on explicit boundary annotations. A dual-level hierarchical difference representation captures temporal irregularities at both fine and coarse scales, while adaptive average pooling preserves essential patterns across variable-length inputs to minimize information loss. Our TDAM-AvgPool model achieves state-of-the-art performance, with an EER of 0.59% on the PartialSpoof dataset and 0.03% on the HAD dataset, which significantly outperforms the existing methods without requiring frame-level supervision.

Figures

Figures reproduced from arXiv: 2507.15101 by the authors.

Figure 1
Figure 1. Comparison of mean value for raw frame-to-frame differences in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison of mean and standard deviation of cosine similarity from [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The architecture of the proposed detection model, TDAM-AvgPool. The model includes a feature extractor with adaptive average pooling and ResNet [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization of ground truth and frame-level weighting maps for [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 31 canonical work pages

  1. [1]

    Asvspoof 2021: Towards spoofed and deepfake speech detection in the wild,

    X. Liu, X. Wang, M. Sahidullah, J. Patino, H. Delgado, T. Kinnunen, M. Todisco, J. Yamagishi, N. Evans, A. Nautsch et al., “Asvspoof 2021: Towards spoofed and deepfake speech detection in the wild,”IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 31, pp. 2507–2522, 2023

    work page 2021

  2. [2]

    Interpretable temporal class activation repre- sentation for audio spoofing detection,

    M. Li and X.-P. Zhang, “Interpretable temporal class activation repre- sentation for audio spoofing detection,” in Interspeech 2024, 2024, pp. 1120–1124

    work page 2024

  3. [3]

    Graph-based spectro- temporal dependency modeling for anti-spoofing,

    F. Chen, S. Deng, T. Zheng, Y . He, and J. Han, “Graph-based spectro- temporal dependency modeling for anti-spoofing,” in ICASSP 2023- 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    work page 2023

  4. [4]

    Graph attention networks for anti-spoofing,

    H. Tak, J. weon Jung, J. Patino, M. Todisco, and N. Evans, “Graph attention networks for anti-spoofing,” in Interspeech 2021 , 2021, pp. 2356–2360

    work page 2021

  5. [5]

    Robust audio anti-spoofing system based on low-frequency sub-band information,

    M. Li and X.-P. Zhang, “Robust audio anti-spoofing system based on low-frequency sub-band information,” in 2023 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA) . IEEE, 2023, pp. 1–5

    work page 2023

  6. [6]

    The partialspoof database and countermeasures for the detection of short fake speech segments embedded in an utterance,

    L. Zhang, X. Wang, E. Cooper, N. Evans, and J. Yamagishi, “The partialspoof database and countermeasures for the detection of short fake speech segments embedded in an utterance,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 31, pp. 813–825, 2022

    work page 2022

  7. [7]

    A survey on speech deepfake detection,

    M. Li, Y . Ahmadiadli, and X.-P. Zhang, “A survey on speech deepfake detection,” ACM Computing Surveys , 2025

    work page 2025

  8. [8]

    Analyzing the impact of splicing artifacts in partially fake speech signals,

    V . Negroni, D. Salvi, P. Bestagini, and S. Tubaro, “Analyzing the impact of splicing artifacts in partially fake speech signals,” in The Automatic Speaker Verification Spoofing Countermeasures Workshop (ASVspoof 2024), 2024, pp. 79–85

    work page 2024

Show all 33 references
  1. [9]

    A comparative study on physical and perceptual features for deepfake audio detection,

    M. Li, Y . Ahmadiadli, and X.-P. Zhang, “A comparative study on physical and perceptual features for deepfake audio detection,” in Proceedings of the 1st International Workshop on Deepfake Detection for Audio Multimedia , 2022, pp. 35–41

    2022

  2. [10]

    Robust ai- synthesized speech detection using feature decomposition learning and synthesizer feature augmentation,

    K. Zhang, Z. Hua, Y . Zhang, Y . Guo, and T. Xiang, “Robust ai- synthesized speech detection using feature decomposition learning and synthesizer feature augmentation,” IEEE Transactions on Information Forensics and Security, 2024

    2024

  3. [11]

    Phoneme-level feature discrepancies: A key to detecting sophisticated speech deepfakes,

    K. Zhang, Z. Hua, R. Lan, Y . Zhang, and Y . Guo, “Phoneme-level feature discrepancies: A key to detecting sophisticated speech deepfakes,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 1, 2025, pp. 1066–1074

    2025

  4. [12]

    Domain generalization via aggregation and separation for audio deepfake detection,

    Y . Xie, H. Cheng, Y . Wang, and L. Ye, “Domain generalization via aggregation and separation for audio deepfake detection,” IEEE Trans- actions on Information Forensics and Security , vol. 19, pp. 344–358, 2023

    2023

  5. [13]

    Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks,

    J.-w. Jung, H.-S. Heo, H. Tak, H.-j. Shim, J. S. Chung, B.-J. Lee, H.-J. Yu, and N. Evans, “Aasist: Audio anti-spoofing using integrated spectro-temporal graph attention networks,” in ICASSP 2022-2022 IEEE international conference on acoustics, speech and signal processing (IC...

    2022

  6. [14]

    Towards end-to-end synthetic speech detection,

    G. Hua, A. B. J. Teoh, and H. Zhang, “Towards end-to-end synthetic speech detection,” IEEE Signal Processing Letters , vol. 28, pp. 1265– 1269, 2021

    2021

  7. [15]

    Detecting synthetic speech manipulation in real audio recordings,

    M. H. Rahman, M. Graciarena, D. Castan, C. Cobo-Kroenke, M. McLaren, and A. Lawson, “Detecting synthetic speech manipulation in real audio recordings,” in 2022 IEEE International Workshop on Information Forensics and Security (WIFS) . IEEE, 2022, pp. 1–6

    2022

  8. [16]

    The vicomtech partial deepfake de- tection and location system for the 2023 add challenge

    J. M. Mart ´ın-Do˜nas and A. ´Alvarez, “The vicomtech partial deepfake de- tection and location system for the 2023 add challenge.” in Proceedings of IJCAI 2023 Workshop on Deepfake Audio Detection and Analysis , 2023, pp. 37–42

    2023

  9. [17]

    Transsionadd: A multi-frame reinforcement based sequence tagging model for audio deepfake detection,

    J. Liu, Z. Su, H. Huang, C. Wan, Q. Wang, J. Hong, B. Tang, and F. Zhu, “Transsionadd: A multi-frame reinforcement based sequence tagging model for audio deepfake detection,” Proceedings of IJCAI 2023 Workshop on Deepfake Audio Detection and Analysis , pp. 113–118, 2023

    2023

  10. [18]

    Convolutional recurrent neural network and multitask learning for manipulation region location

    K. Li, X.-M. Zeng, J.-T. Zhang, and Y . Song, “Convolutional recurrent neural network and multitask learning for manipulation region location.” in Proceedings of IJCAI 2023 Workshop on Deepfake Audio Detection and Analysis, 2023, pp. 18–22

    2023

  11. [19]

    Multi-grained backend fusion for manipulation region location of partially fake audio

    J. Li, L. Li, M. Luo, X. Wang, S. Qiao, and Y . Zhou, “Multi-grained backend fusion for manipulation region location of partially fake audio.” in Proceedings of IJCAI 2023 Workshop on Deepfake Audio Detection and Analysis, 2023, pp. 43–48

    2023

  12. [20]

    Multi-task learning in utterance-level and segmental-level spoof detection,

    L. Zhang, X. Wang, E. Cooper, and J. Yamagishi, “Multi-task learning in utterance-level and segmental-level spoof detection,” in 2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge, 2021, pp. 9–15

    2021

  13. [21]

    Fake audio detection based on unsupervised pretraining models,

    Z. Lv, S. Zhang, K. Tang, and P. Hu, “Fake audio detection based on unsupervised pretraining models,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2022, pp. 9231–9235

    2022

  14. [22]

    Integrating frame-level boundary detection and deepfake detection for locating manipulated regions in partially spoofed audio forgery attacks,

    Z. Cai and M. Li, “Integrating frame-level boundary detection and deepfake detection for locating manipulated regions in partially spoofed audio forgery attacks,” Computer Speech & Language , vol. 85, p. 101597, 2024

    2024

  15. [23]

    How do neural spoofing countermeasures detect partially spoofed audio?

    T. Liu, L. Zhang, R. K. Das, Y . Ma, R. Tao, and H. Li, “How do neural spoofing countermeasures detect partially spoofed audio?” in Interspeech 2024, 2024, pp. 1105–1109

    2024

  16. [24]

    Spatiotemporal inconsistency learning for deepfake video detection,

    Z. Gu, Y . Chen, T. Yao, S. Ding, J. Li, F. Huang, and L. Ma, “Spatiotemporal inconsistency learning for deepfake video detection,” in Proceedings of the 29th ACM international conference on multimedia , 2021, pp. 3473–3481

    2021

  17. [25]

    Asvspoof 2019: Future horizons in spoofed and fake audio detection,

    M. Todisco, X. Wang, V . Vestman, M. Sahidullah, H. Delgado, A. Nautsch, J. Yamagishi, N. Evans, T. H. Kinnunen, and K. A. Lee, “Asvspoof 2019: Future horizons in spoofed and fake audio detection,” in Interspeech 2019, 2019, pp. 1008–1012

    2019

  18. [26]

    Xls-r: Self-supervised cross-lingual speech representation learning at scale,

    A. Babu, C. Wang, A. Tjandra, K. Lakhotia, Q. Xu, N. Goyal, K. Singh, P. von Platen, Y . Saraf, J. Pino, A. Baevski, A. Conneau, and M. Auli, “Xls-r: Self-supervised cross-lingual speech representation learning at scale,” in Interspeech 2022, 2022, pp. 2278–2282

    2022

  19. [27]

    Half- truth: A partially fake audio detection dataset,

    J. Yi, Y . Bai, J. Tao, H. Ma, Z. Tian, C. Wang, T. Wang, and R. Fu, “Half- truth: A partially fake audio detection dataset,” in Interspeech 2021 , 2021, pp. 1654–1658

    2021

  20. [28]

    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

  21. [29]

    Frame-to-utterance convergence: A spectra-temporal approach for unified spoofing detection,

    A. Khan, K. M. Malik, and S. Nawaz, “Frame-to-utterance convergence: A spectra-temporal approach for unified spoofing detection,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 10 761–10 765

    2024

  22. [30]

    Local self-attention-based hybrid multiple instance learning for partial spoof speech detection,

    Y . Zhu, Y . Chen, Z. Zhao, X. Liu, and J. Guo, “Local self-attention-based hybrid multiple instance learning for partial spoof speech detection,” ACM Transactions on Intelligent Systems and Technology, vol. 14, no. 5, pp. 1–18, 2023

    2023

  23. [31]

    Weakly-supervised audio temporal forgery localization via progressive audio-language co- learning network,

    J. Wu, W. Xu, W. Lu, X. Luo, R. Yang, and S. Guo, “Weakly-supervised audio temporal forgery localization via progressive audio-language co- learning network,” arXiv preprint arXiv:2505.01880 , 2025

    2025 arXiv

  24. [32]

    Coarse-to-fine proposal refinement framework for audio temporal forgery detection and localization,

    J. Wu, W. Lu, X. Luo, R. Yang, Q. Wang, and X. Cao, “Coarse-to-fine proposal refinement framework for audio temporal forgery detection and localization,” in Proceedings of the 32nd ACM International Conference on Multimedia, 2024, pp. 7395–7403

    2024

  25. [33]

    Wavlm and omni-scale cnns: Enhancing boundary detection in partially spoofed audio,

    M. Li and Z. Huang, “Wavlm and omni-scale cnns: Enhancing boundary detection in partially spoofed audio,” in 2024 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 2024, pp. 1–5

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.