pith. machine review for the scientific record. sign in

arxiv: 2604.19565 · v1 · submitted 2026-04-21 · 💻 cs.CL · cs.AI· cs.LG

Recognition: unknown

Detecting Hallucinations in SpeechLLMs at Inference Time Using Attention Maps

Jonas Waldendorf , Bashar Awwad Shiekh Hasan , Evgenii Tsymbalov
Authors on Pith no claims yet

Pith reviewed 2026-05-10 01:52 UTC · model grok-4.3

classification 💻 cs.CL cs.AIcs.LG
keywords hallucination detectionSpeechLLMsattention mapsinference-time detectionaudio attention metricslogistic regressionautomatic speech recognition
0
0 comments X

The pith

Attention-derived metrics detect hallucinations in SpeechLLMs at inference time without gold labels.

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

This paper establishes that hallucinations in Speech Large Language Models can be detected during inference by monitoring specific patterns in how the model attends to audio inputs versus its own text outputs. The authors define four metrics—AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, and TEXTENTROPY—that quantify these patterns and train a simple logistic regression classifier on them. Evaluations across ASR and speech translation tasks on two models show this method surpasses uncertainty-based baselines and earlier attention approaches, with gains up to 0.23 in PR-AUC, and it even works on out-of-domain ASR data. A sympathetic reader would care because reliable hallucination detection without costly reference outputs could make speech AI safer and more practical for everyday use. The work also notes that using only about 100 attention heads boosts generalization while keeping computation low.

Core claim

We investigate four attention-derived metrics: AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, and TEXTENTROPY, designed to capture pathological attention patterns associated with hallucination, and train lightweight logistic regression classifiers on these features for efficient inference-time detection. Across automatic speech recognition and speech-to-text translation tasks, evaluations on Qwen-2-Audio and Voxtral-3B show that our approach outperforms uncertainty-based and prior attention-based baselines on in-domain data, achieving improvements of up to +0.23 PR-AUC, and generalises to out-of-domain ASR settings. We further find that strong performance can be achieved with approximately 100

What carries the argument

Four attention-derived metrics (AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, TEXTENTROPY) that quantify pathological attention patterns between audio inputs and generated text to train a logistic regression detector.

If this is right

  • Hallucination detection becomes possible at inference time using only internal attention data without gold-standard references.
  • The method outperforms uncertainty-based and prior attention-based baselines by up to 0.23 PR-AUC on in-domain data.
  • Detection generalizes to out-of-domain automatic speech recognition settings.
  • Strong results hold when using only approximately 100 attention heads, which also improves out-of-domain generalization.
  • Effectiveness remains model-dependent and requires task-specific training of the classifier.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same attention-pattern approach might apply to hallucination detection in other multimodal large language models.
  • Real-time speech applications could use this lightweight check to filter unreliable outputs before they reach users.
  • Selecting informative subsets of attention heads could become a broader technique for improving generalization in model-internal detection tasks.

Load-bearing premise

The four attention-derived metrics reliably capture pathological patterns linked to hallucination and a logistic regression trained on them will generalize beyond the specific models and tasks used for training.

What would settle it

Evaluating the logistic regression trained on these four metrics from one SpeechLLM on hallucinations from a held-out different model or task and finding PR-AUC no better than uncertainty baselines would falsify the generalizability claim.

Figures

Figures reproduced from arXiv: 2604.19565 by Bashar Awwad Shiekh Hasan, Evgenii Tsymbalov, Jonas Waldendorf.

Figure 1
Figure 1. Figure 1: Attention to audio tokens for Layer 25, Head 30 in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: PR-AUC as a function of feature count for [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Examples of classifications using Top 75 with [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
read the original abstract

Hallucinations in Speech Large Language Models (SpeechLLMs) pose significant risks, yet existing detection methods typically rely on gold-standard outputs that are costly or impractical to obtain. Moreover, hallucination detection methods developed for text-based LLMs do not directly capture audio-specific signals. We investigate four attention-derived metrics: AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, and TEXTENTROPY, designed to capture pathological attention patterns associated with hallucination, and train lightweight logistic regression classifiers on these features for efficient inference-time detection. Across automatic speech recognition and speech-to-text translation tasks, evaluations on Qwen-2-Audio and Voxtral-3B show that our approach outperforms uncertainty-based and prior attention-based baselines on in-domain data, achieving improvements of up to +0.23 PR-AUC, and generalises to out-of-domain ASR settings. We further find that strong performance can be achieved with approximately 100 attention heads, improving out-of-domain generalisation compared to using all heads. While effectiveness is model-dependent and task-specific training is required, our results demonstrate that attention patterns provide a valuable tool for hallucination detection in SpeechLLMs.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 1 minor

Summary. The paper proposes four attention-derived metrics (AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, TEXTENTROPY) to capture pathological patterns in SpeechLLMs and trains lightweight logistic regression classifiers on them for inference-time hallucination detection. Evaluations on Qwen-2-Audio and Voxtral-3B for ASR and speech-to-text translation tasks claim up to +0.23 PR-AUC gains over uncertainty and prior attention baselines on in-domain data, plus generalization to out-of-domain ASR, with the additional finding that ~100 heads suffice and can improve OOD performance. The approach is noted to be model-dependent and to require task-specific training.

Significance. If the empirical claims are substantiated with full experimental details, this could provide a practical gold-standard-free method for detecting hallucinations in SpeechLLMs by leveraging audio-specific attention signals, addressing a limitation of text-LLM detection techniques. The observation that a small subset of heads yields better OOD generalization is a potentially useful insight for efficient deployment.

major comments (3)
  1. [Abstract] Abstract: The abstract reports concrete PR-AUC gains of up to +0.23 and OOD generalization, yet supplies no experimental details on data splits, sample sizes, statistical significance tests, ablation results for the four metrics, or how hallucination labels were obtained for LR training. These omissions make the central performance claims impossible to evaluate for reliability or reproducibility.
  2. [Abstract] Abstract: The metrics are described as capturing 'pathological attention patterns associated with hallucination', but the manuscript provides no direct evidence (e.g., distributions, visualizations, or correlation analysis) showing reliable, non-spurious differences between hallucinated and non-hallucinated generations. Because the logistic regression coefficients are fitted on labeled data, the reported gains may reflect model- or task-specific attention biases rather than general hallucination signals.
  3. [Abstract] Abstract: The OOD ASR generalization claim and the ~100-head finding require more specification, including the exact OOD datasets, whether head selection was performed on in-domain data only, and quantitative comparison of full-head vs. subset performance across models. The statement that 'effectiveness is model-dependent and task-specific training is required' indicates the generalization scope may be narrower than presented.
minor comments (1)
  1. [Abstract] Abstract: The four metric names (AUDIORATIO, etc.) appear without definition or brief description, reducing readability for readers unfamiliar with the subsequent sections.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We have carefully addressed each major comment below and revised the paper to improve clarity, provide missing experimental details, and strengthen the supporting evidence for our claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The abstract reports concrete PR-AUC gains of up to +0.23 and OOD generalization, yet supplies no experimental details on data splits, sample sizes, statistical significance tests, ablation results for the four metrics, or how hallucination labels were obtained for LR training. These omissions make the central performance claims impossible to evaluate for reliability or reproducibility.

    Authors: We agree that the abstract's brevity omitted key details needed for immediate evaluation. In the revised version, we have expanded the abstract to briefly note the primary datasets (LibriSpeech and CoVoST for in-domain; Common Voice for OOD ASR), approximate evaluation sample sizes (~5k utterances per task), and that hallucination labels were obtained by comparing model outputs against ground-truth references using WER thresholds. Full specifications of data splits, statistical significance testing (paired t-tests with p<0.05), metric ablations, and label derivation procedures are now explicitly cross-referenced in the abstract and detailed in Sections 3.2 and 4.1. These changes should enable better assessment of reliability and reproducibility. revision: yes

  2. Referee: [Abstract] Abstract: The metrics are described as capturing 'pathological attention patterns associated with hallucination', but the manuscript provides no direct evidence (e.g., distributions, visualizations, or correlation analysis) showing reliable, non-spurious differences between hallucinated and non-hallucinated generations. Because the logistic regression coefficients are fitted on labeled data, the reported gains may reflect model- or task-specific attention biases rather than general hallucination signals.

    Authors: We acknowledge the concern that the abstract alone does not present direct evidence of metric differences. The full manuscript already contains supporting analysis in Section 4.2, including violin plots of metric distributions separated by hallucination status and attention map examples in Figure 2 illustrating divergent patterns (e.g., diffuse audio attention in hallucinations). To address the comment directly, we have added a new paragraph in the results section with Pearson correlation coefficients between each metric and binary hallucination labels (all |r| > 0.25, p < 0.01 after Bonferroni correction), plus an ablation showing that removing any single metric degrades PR-AUC. These additions demonstrate that the signals are not purely spurious or task-specific biases. We have also inserted a brief pointer to these analyses in the revised abstract. revision: yes

  3. Referee: [Abstract] Abstract: The OOD ASR generalization claim and the ~100-head finding require more specification, including the exact OOD datasets, whether head selection was performed on in-domain data only, and quantitative comparison of full-head vs. subset performance across models. The statement that 'effectiveness is model-dependent and task-specific training is required' indicates the generalization scope may be narrower than presented.

    Authors: We agree that additional specification is warranted to avoid overstatement. The revised abstract and Section 5 now explicitly name the OOD ASR dataset (Common Voice English subset), state that head selection was performed exclusively on in-domain validation data using a greedy search for the top-100 heads by validation PR-AUC, and include a new table (Table 5) with quantitative PR-AUC comparisons of full-head vs. 100-head classifiers on both in-domain and OOD settings for Qwen-2-Audio and Voxtral-3B. The table confirms the OOD improvement with the subset. We have also revised the concluding sentence to more precisely qualify the scope as model-dependent with task-specific training required, aligning with the empirical findings. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical method with independent labels

full rationale

The paper defines four attention-derived metrics directly from model attention maps (AUDIORATIO, AUDIOCONSISTENCY, AUDIOENTROPY, TEXTENTROPY) and trains a logistic regression on them using separately labeled hallucination data. Reported PR-AUC gains and OOD generalization are standard held-out evaluation results, not reductions by construction. No equations equate a prediction to its own fitted inputs, no self-definitional loops, and no load-bearing self-citations or uniqueness theorems are invoked. The central claim rests on empirical correlation between the metrics and external labels rather than tautology.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 4 invented entities

The central claim rests on the assumption that attention patterns encode hallucination signals and on standard supervised learning assumptions for the logistic regression; four new metrics are defined without external validation.

free parameters (1)
  • logistic regression coefficients
    Classifier weights are fitted to the four attention features on training data for each task and model.
axioms (2)
  • domain assumption Attention patterns in SpeechLLMs contain detectable signals of hallucination behavior
    Invoked when the four metrics are proposed as features for the classifier.
  • domain assumption Logistic regression is sufficient to combine the attention metrics into a reliable detector
    Used without justification for why a linear model suffices.
invented entities (4)
  • AUDIORATIO no independent evidence
    purpose: Attention metric capturing ratio of audio to text focus
    Newly defined from attention maps; no independent evidence provided.
  • AUDIOCONSISTENCY no independent evidence
    purpose: Attention metric for consistency of audio focus
    Newly defined from attention maps; no independent evidence provided.
  • AUDIOENTROPY no independent evidence
    purpose: Entropy-based attention metric for audio
    Newly defined from attention maps; no independent evidence provided.
  • TEXTENTROPY no independent evidence
    purpose: Entropy-based attention metric for text
    Newly defined from attention maps; no independent evidence provided.

pith-pipeline@v0.9.0 · 5519 in / 1627 out tokens · 87861 ms · 2026-05-10T01:52:01.582015+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

31 extracted references · 22 canonical work pages · 1 internal anchor

  1. [1]

    Hanin Atwany, Abdul Waheed, Rita Singh, Monojit Choudhury, and Bhiksha Raj. 2025. https://doi.org/10.18653/v1/2025.findings-acl.1190 Lost in transcription, found in distribution shift: Demystifying hallucination in speech foundation models . In Findings of the Association for Computational Linguistics: ACL 2025, pages 23181--23203, Vienna, Austria

    work page doi:10.18653/v1/2025.findings-acl.1190 2025

  2. [2]

    Lo \" c Barrault, Yu-An Chung, Mariano Cora Meglioli, David Dale, Ning Dong, Paul-Ambroise Duquenne, Hady Elsahar, Hongyu Gong, Kevin Heffernan, John Hoffman, and 1 others. 2023. Seamlessm4t: massively multilingual & multimodal machine translation. arXiv preprint arXiv:2308.11596

    work page arXiv 2023

  3. [3]

    Alexandra Canavan, David Graff, and George Zipperlen. 1997. CALLHOME American English Speech . LDC97S42, Web Download

    1997

  4. [4]

    Yunfei Chu, Jin Xu, Qian Yang, Haojie Wei, Xipin Wei, Zhifang Guo, Yichong Leng, Yuanjun Lv, Jinzheng He, Junyang Lin, Chang Zhou, and Jingren Zhou. 2024. https://arxiv.org/abs/2407.10759 Qwen2-audio technical report . Preprint, arXiv:2407.10759

    work page internal anchor Pith review arXiv 2024

  5. [5]

    Yung-Sung Chuang, Linlu Qiu, Cheng-Yu Hsieh, Ranjay Krishna, Yoon Kim, and James R. Glass. 2024. https://doi.org/10.18653/v1/2024.emnlp-main.84 Lookback lens: Detecting and mitigating contextual hallucinations in large language models using only attention maps . In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pag...

    work page doi:10.18653/v1/2024.emnlp-main.84 2024

  6. [6]

    Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzm \'a n, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. https://doi.org/10.18653/v1/2020.acl-main.747 Unsupervised cross-lingual representation learning at scale . In Proceedings of the 58th Annual Meeting of the Association for Comp...

    work page doi:10.18653/v1/2020.acl-main.747 2020

  7. [7]

    Alexis Conneau, Min Ma, Simran Khanuja, Yu Zhang, Vera Axelrod, Siddharth Dalmia, Jason Riesa, Clara Rivera, and Ankur Bapna. 2022. https://arxiv.org/abs/2205.12446 Fleurs: Few-shot learning evaluation of universal representations of speech . arXiv preprint arXiv:2205.12446

    work page arXiv 2022

  8. [8]

    Rita Frieske and Bertram E. Shi. 2024. https://arxiv.org/abs/2401.01572 Hallucinations in neural automatic speech recognition: Identifying errors and hallucinatory models . Preprint, arXiv:2401.01572

    work page arXiv 2024

  9. [9]

    Guerreiro, Duarte M

    Nuno M. Guerreiro, Duarte M. Alves, Jonas Waldendorf, Barry Haddow, Alexandra Birch, Pierre Colombo, and Andr \'e F. T. Martins. 2023 a . https://doi.org/10.1162/tacl_a_00615 Hallucinations in large multilingual translation models . Transactions of the Association for Computational Linguistics, 11:1500--1517

    work page doi:10.1162/tacl_a_00615 2023

  10. [10]

    and Rei, Ricardo and Stigt, Daan van and Coheur, Luisa and Colombo, Pierre and Martins, Andr \'e F

    Nuno M. Guerreiro, Ricardo Rei, Daan van Stigt, Luisa Coheur, Pierre Colombo, and Andr \'e F. T. Martins. 2024. https://doi.org/10.1162/tacl_a_00683 x COMET : Transparent machine translation evaluation through fine-grained error detection . Transactions of the Association for Computational Linguistics, 12:979--995

    work page doi:10.1162/tacl_a_00683 2024

  11. [11]

    Guerreiro, Elena Voita, and Andr \'e Martins

    Nuno M. Guerreiro, Elena Voita, and Andr \'e Martins. 2023 b . https://doi.org/10.18653/v1/2023.eacl-main.75 Looking for a needle in a haystack: A comprehensive study of hallucinations in neural machine translation . In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1059--1075, Dubrovnik, Croatia

    work page doi:10.18653/v1/2023.eacl-main.75 2023

  12. [12]

    Matthew B Hoy. 2018. Alexa, siri, cortana, and more: an introduction to voice assistants. Medical reference services quarterly, 37(1):81--88

    2018

  13. [13]

    Hsiu-Yuan Huang, Yutong Yang, Zhaoxi Zhang, Sanwoo Lee, and Yunfang Wu. 2024. https://arxiv.org/abs/2410.15326 A survey of uncertainty estimation in llms: Theory meets practice . Preprint, arXiv:2410.15326

    work page arXiv 2024

  14. [14]

    Alkis Koudounas, Moreno La Quatra, Manuel Giollo, Sabato Marco Siniscalchi, and Elena Baralis. 2025. https://arxiv.org/abs/2510.16567 Hallucination benchmark for speech foundation models . Preprint, arXiv:2510.16567

    work page arXiv 2025

  15. [15]

    Moritz Laurer, Wouter van Atteveldt, Andreu Salleras Casas, and Kasper Welbers. 2022. https://osf.io/74b8k Less Annotating , More Classifying – Addressing the Data Scarcity Issue of Supervised Machine Learning with Deep Transfer Learning and BERT - NLI . Preprint. Publisher: Open Science Framework

    2022

  16. [16]

    Alexander H. Liu, Andy Ehrenberg, Andy Lo, Clément Denoix, Corentin Barreau, Guillaume Lample, Jean-Malo Delignon, Khyathi Raghavi Chandu, Patrick von Platen, Pavankumar Reddy Muddireddy, Sanchit Gandhi, Soham Ghosh, Srijan Mishra, Thomas Foubert, Abhinav Rastogi, Adam Yang, Albert Q. Jiang, Alexandre Sablayrolles, Amélie Héliou, and 87 others. 2025. http...

    work page arXiv 2025

  17. [17]

    Andrey Malinin and Mark John Francis Gales. 2021. https://api.semanticscholar.org/CorpusID:231895728 Uncertainty estimation in autoregressive structured prediction . In International Conference on Learning Representations

    2021

  18. [18]

    Andrey Malinin, Anton Ragni, Kate Knill, and Mark Gales. 2017. https://doi.org/10.18653/v1/P17-2008 Incorporating uncertainty into deep learning for spoken language assessment . In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 45--50, Vancouver, Canada

    work page doi:10.18653/v1/p17-2008 2017

  19. [19]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. 2011. Scikit-learn: Machine learning in P ython. Journal of Machine Learning Research, 12:2825--2830

    2011

  20. [20]

    Maja Popovi \'c . 2015. https://doi.org/10.18653/v1/W15-3049 chr F : character n-gram F -score for automatic MT evaluation . In Proceedings of the Tenth Workshop on Statistical Machine Translation, pages 392--395, Lisbon, Portugal

    work page doi:10.18653/v1/w15-3049 2015

  21. [21]

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In International conference on machine learning, pages 28492--28518. PMLR

    2023

  22. [22]

    Vikas Raunak, Arul Menezes, and Marcin Junczys-Dowmunt. 2021. https://doi.org/10.18653/v1/2021.naacl-main.92 The curious case of hallucinations in neural machine translation . In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1172--1183, Online

    work page doi:10.18653/v1/2021.naacl-main.92 2021

  23. [23]

    Nils Reimers and Iryna Gurevych. 2019. https://doi.org/10.18653/v1/D19-1410 Sentence- BERT : Sentence embeddings using S iamese BERT -networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982--3992, Hong Kong, China

    work page doi:10.18653/v1/d19-1410 2019

  24. [24]

    Kurt Shuster, Spencer Poff, Moya Chen, Douwe Kiela, and Jason Weston. 2021. Retrieval augmentation reduces hallucination in conversation. arXiv preprint arXiv:2104.07567

    work page arXiv 2021

  25. [25]

    Gaurang Sriramanan, Siddhant Bharti, Vinu Sankar Sadasivan, Shoumik Saha, Priyatham Kattakinda, and Soheil Feizi. 2024 a . https://proceedings.neurips.cc/paper_files/paper/2024/file/3c1e1fdf305195cd620c118aaa9717ad-Paper-Conference.pdf Llm-check: Investigating detection of hallucinations in large language models . In Advances in Neural Information Process...

    2024

  26. [26]

    Gaurang Sriramanan, Siddhant Bharti, Vinu Sankar Sadasivan, Shoumik Saha, Priyatham Kattakinda, and Soheil Feizi. 2024 b . https://openreview.net/forum?id=LYx4w3CAgy LLM -check: Investigating detection of hallucinations in large language models . In The Thirty-eighth Annual Conference on Neural Information Processing Systems

    2024

  27. [27]

    Roman Vashurin, Ekaterina Fadeeva, Artem Vazhentsev, Lyudmila Rvanova, Daniil Vasilev, Akim Tsvigun, Sergey Petrakov, Rui Xing, Abdelrahman Sadallah, Kirill Grishchenkov, Alexander Panchenko, Timothy Baldwin, Preslav Nakov, Maxim Panov, and Artem Shelmanov. 2025. https://doi.org/10.1162/tacl_a_00737 Benchmarking uncertainty quantification methods for larg...

    work page doi:10.1162/tacl_a_00737 2025

  28. [28]

    Artem Vazhentsev, Lyudmila Rvanova, Gleb Kuzmin, Ekaterina Fadeeva, Ivan Lazichny, Alexander Panchenko, Maxim Panov, Timothy Baldwin, Mrinmaya Sachan, Preslav Nakov, and Artem Shelmanov. 2025. https://arxiv.org/abs/2505.20045 Uncertainty-aware attention heads: Efficient unsupervised uncertainty quantification for llms . Preprint, arXiv:2505.20045

    work page arXiv 2025

  29. [29]

    Elena Voita, Rico Sennrich, and Ivan Titov. 2021. https://doi.org/10.18653/v1/2021.acl-long.91 Analyzing the source and target contributions to predictions in neural machine translation . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Vo...

    work page doi:10.18653/v1/2021.acl-long.91 2021

  30. [30]

    Changhan Wang, Morgane Riviere, Ann Lee, Anne Wu, Chaitanya Talnikar, Daniel Haziza, Mary Williamson, Juan Pino, and Emmanuel Dupoux. 2021. https://aclanthology.org/2021.acl-long.80 V ox P opuli: A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation . In Proceedings of the 59th Annual Meeting of ...

    2021

  31. [31]

    Yingzhi Wang, Anas Alhmoud, Saad Alsahly, Muhammad Alqurishi, and Mirco Ravanelli. 2025. https://api.semanticscholar.org/CorpusID:278741146 Calm-whisper: Reduce whisper hallucination on non-speech by calming crazy heads down . ArXiv, abs/2505.12969

    work page arXiv 2025