Pith. sign in

REVIEW 5 minor 50 references

Gradient-Based Speech-to-Text Alignment for Any ASR Model: From CTC to Speech LLMs

T0 review · 0 major / 5 minor · reviewed 2026-07-10 · grok-4.5

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

Pith's one-line read A single gradient-of-token-probability signal yields usable word boundaries for every differentiable ASR model, including speech LLMs.

desk verdict Solid, carefully scoped empirical paper: one training-free gradient procedure works as a forced aligner across CTC, transducers, AED and speech LLMs, with thorough evaluation and honest hedging. read the letter →

arxiv 2607.06831 v1 pith:H3U2GBPV submitted 2026-07-07 cs.CL cs.AIcs.CVcs.LG

Albert Zeyer , Ralf Schlüter , Hermann Ney This is my paper · ORCID
classification cs.CLcs.AIcs.CVcs.LG
keywords speech-to-textalignmentforcedgradient-basedattributionautomaticspeechrecognitionlanguagemodelsCTCtransducersattention-basedencoder-decoder
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Speech-to-text alignment tells you when each word in a transcript begins and ends in the audio. Classic models give you that timing by construction; modern attention models and speech language models usually do not, and people read timings off attention weights on a coarse encoder grid. This paper shows that you can instead take the gradient of each teacher-forced token log-probability with respect to the raw input, turn the gradient into a per-frame saliency matrix, and recover word boundaries with one dynamic-programming pass. The procedure needs no training, no extra heads, and no model changes. On sixteen models spanning CTC, transducers, attention-based encoder-decoders and speech LLMs, evaluated on both read and spontaneous speech, the gradient alignment is usable everywhere, trails a strong native aligner only modestly, and beats the native signal precisely where that signal is weak (streaming models and one speech LLM). Its main cost is one backward pass per token.

What carries the argument

Input-gradient saliency matrix: for each transcript token, take the log L2-norm of the gradient of its teacher-forced log-probability w.r.t. the input frames, optionally energy-weight it, apply a log-softmax over time, then decode the token-by-frame matrix with a word-level blank topology via time-synchronous Viterbi. This single signal replaces model-specific posteriors or attention heads and lives on the fine input grid rather than the coarse encoder grid.

What would settle it

On a held-out spontaneous-speech set, for any of the streaming models or Canary-Qwen, replace the gradient scores with random or constant saliency of the same shape and show that the dynamic program still recovers word-boundary errors within a few tens of milliseconds of the reported figures; if it does, the claimed saliency content is not doing the work.

Watch

Extended reading notes

Core claim

A training-free procedure—gradient of each teacher-forced token log-probability with respect to the input, reduced to per-frame saliency and decoded by a single dynamic-programming pass—produces a usable word-boundary alignment for every differentiable ASR family tested (CTC, transducers, AED, speech LLMs) and is better than the model’s own native or attention alignment precisely where that native signal is weak.

Load-bearing premise

That the magnitude of the input gradient of a single token’s log-probability is already a clean enough local saliency signal that a simple energy-weighted dynamic program can recover accurate word boundaries.

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

0 major / 5 minor

Summary. The paper proposes a training-free gradient-based forced alignment that applies to any differentiable ASR model. For each teacher-forced token it computes the gradient of the token log-probability with respect to the input, reduces it to a per-frame saliency (log L2-norm), optionally energy-weights the scores, and recovers word boundaries by a single time-synchronous Viterbi pass over a word-level blank topology. The method is shown to work for CTC (via prefix scores), transducers (RNN-T and TDT), attention-based encoder-decoders, and speech LLMs. Evaluation covers sixteen models on TIMIT and Buckeye, comparing gradient alignment against each model’s own native posterior or attention-based alignment (and against MFA as a classical reference). The central empirical claim is that the gradient yields a usable alignment for every family, is usually a little behind a strong native aligner, but is better precisely where the native signal is weak (streaming models, Canary-Qwen), and that its main cost is one backward pass per token.

Significance. If the result holds, the paper supplies a single, model-agnostic, training-free procedure that produces word timings for every current ASR family, including speech LLMs that otherwise lack a built-in aligner. The evaluation is unusually thorough (sixteen models, two corpora, reference and hypothesis modes, extensive ablations on reduction norms, blank schemes, energy weighting, encoder depth, tokenization, and compute cost) and the claims are carefully hedged. Public source code is promised, which strengthens reproducibility. The work is therefore a useful reference result and analysis tool even if practitioners continue to prefer cheaper native aligners when they are strong.

minor comments (5)
  1. Table I caption and surrounding text should state more explicitly that the starred Whisper-large-v3 row uses an intermediate encoder depth chosen on development data; a casual reader may otherwise treat it as the default input-gradient result.
  2. Section II, Eq. (1)–(4): a short remark on numerical stability of the log-norm and of the energy-aware blank (especially when σ_t is near zero) would help implementers.
  3. Figure 1 is dense; adding a brief legend for the colour scale of the saliency matrices would improve readability.
  4. The compute-cost table (Table XII) reports only a subset of models; a one-sentence note that the RTF ordering is expected to be similar for the remaining families would be useful.
  5. A few typographical inconsistencies remain (e.g., “V oxtral”, “Schl ¨uter”); a final pass would clean them.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: method is a constructive saliency+DP procedure evaluated against external gold boundaries and independent native/attention baselines; only minor non-load-bearing self-citation of prior gradient applications.

full rationale

The paper defines gradient saliency Gs,t of teacher-forced token log-probabilities (Eq. 1), reduces it via log-softmax and optional energy weighting (Eq. 3), and decodes word boundaries by a single time-synchronous Viterbi DP over an explicit FSA topology (Eq. 2, Fig. 2). This is a self-contained constructive procedure, not a derivation that reduces to its own inputs. All reported claims are empirical comparisons of the resulting WBE / ≤50 ms accuracy against (a) gold word boundaries on TIMIT-test and Buckeye and (b) each model’s independently computed native posteriors or attention DTW/self-attention, across 16 models. Hyper-parameters (ρ, blank schemes eta t, p-norms, topology) are openly ablated (Tables V–VIII) rather than fitted to the target metric and then re-reported as predictions. Head selection for attention baselines uses a labeled development set and is disclosed. The sole self-citations ([23] and related) merely note prior applications of the same gradient idea to AED models and the observation that encoders can reverse time; they are not invoked as uniqueness theorems, forced ansätze, or load-bearing premises that close the argument. No self-definitional loop, fitted-input-as-prediction, or renaming of a known result appears. The central claim therefore stands or falls on the external numbers, not on circular construction.

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

The work is empirical and method-driven. It rests on standard automatic-differentiation and dynamic-programming machinery plus a small set of hand-chosen decoding hyper-parameters. No new physical entities or free constants are introduced; the free parameters are ordinary algorithmic knobs whose sensitivity is ablated.

free parameters (3)
  • energy-weighting exponent ρ = 0.5
    Default ρ=0.5 multiplies log-energy into the saliency before log-softmax; swept in Table VII.
  • blank-score hyper-parameters (γ, κ, λ) = λ≈2 or κ≈1
    Constant, z-score or energy-aware blank level used inside the DP; defaults chosen by ablation on Buckeye (Table VI).
  • attention-head selection (top-8 by WBE) = top 8 heads
    For AED/LLM baselines the best heads are chosen on TIMIT development gold; not used for the gradient method itself.
assumptions (3)
  • domain assumption The L2-norm (or similar p-norm) of the input gradient of a teacher-forced token log-probability is a temporally meaningful saliency for that token.
    Core modeling choice of Section II; supported by prior saliency literature but not derived.
  • domain assumption Word-level topology (blanks only at word boundaries) plus energy-aware blank scoring is an adequate alignment label topology for evaluation of word boundaries.
    Stated in Section II; alternatives (full CTC-like, no-interior-silence) are ablated.
  • standard math Prefix scores for CTC and transducers correctly recover the teacher-forced token log-probability needed for the gradient.
    Standard dynamic-programming identity cited via Hori et al. [25].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gradient-Based Speech-to-Text Alignment for Any ASR Model: From CTC to Speech LLMs." pith.science (2026). https://pith.science/paper/H3U2GBPV

@misc{pith2026260706831,
  author       = {Pith},
  title        = {Pith review of: Gradient-Based Speech-to-Text Alignment for Any ASR Model: From CTC to Speech LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H3U2GBPV}},
  note         = {Machine review of arXiv:2607.06831}
}
read the original abstract

Speech-to-text alignment means finding the temporal boundaries of each word in the audio. Some models provide such an alignment directly and others do not. Connectionist temporal classification (CTC) and transducer models have an alignment by construction, whereas attention-based encoder-decoders (AED) and speech large language models (LLMs) do not, and their word timings are usually read off the attention weights instead. All of these signals live on the encoder frame grid, which bounds their temporal precision. We study a generic gradient-based alignment that applies to any differentiable ASR model. We take the gradient of each teacher-forced token log probability with respect to the input, reduce it to a per-frame saliency, and decode the resulting matrix into word boundaries with a single dynamic-programming pass. The method needs no training, no model modification and no alignment heads, works across all model families including the speech LLMs, and aligns on the input grid rather than on the coarser encoder grid. We evaluate it on sixteen models from four families, on read (TIMIT) and spontaneous (Buckeye) speech, each against the model's own native or attention-based alignment. We find that the gradient yields a usable alignment for every model, that it is usually somewhat behind a strong native aligner but better where the native alignment is weak, as for the streaming models, and that its main disadvantage is the cost of one backward pass per token.

Figures

Figures reproduced from arXiv: 2607.06831 by the authors.

Figure 1
Figure 1. Posteriors log  pt(y=as | x T ′ 1 )  s,t ∈ R S×T , gradient scores log softmaxT (G) ∈ R S×T , and log self-attention weights (in R S×T ), all without energy weighting here, each with word boundaries, in comparison to the reference segmen￾tation (silence in white, words in blue, with word boundaries). ϵ between two equal labels as = as+1), and a variant with no interior silence keeps only the leading and trailing b… view at source ↗
Figure 2
Figure 2. Decoding the saliency into an alignment, for two words [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages

  1. [1]

    The application of hidden Markov models in speech recognition,

    M. Gales and S. Young, “The application of hidden Markov models in speech recognition,”Found. Trends Signal Process., vol. 1, no. 3, pp. 195–304, Jan. 2008

    work page 2008

  2. [2]

    A tutorial on hidden Markov models and selected applications in speech recognition,

    L. R. Rabiner, “A tutorial on hidden Markov models and selected applications in speech recognition,”Proc. of the IEEE, vol. 77, no. 2, pp. 257–286, 1989

    work page 1989

  3. [3]

    Montreal Forced Aligner: Trainable text-speech alignment using Kaldi,

    M. McAuliffe, M. Socolof, S. Mihuc, M. Wagner, and M. Sonderegger, “Montreal Forced Aligner: Trainable text-speech alignment using Kaldi,” inProc. Interspeech, 2017, pp. 498–502

    work page 2017

  4. [4]

    Less peaky and more accurate CTC forced alignment by label priors,

    R. Huang, X. Zhang, Z. Ni, L. Sun, M. Hira, J. Hwang, V . Manohar, V . Pratap, M. Wiesner, S. Watanabe, D. Povey, and S. Khudanpur, “Less peaky and more accurate CTC forced alignment by label priors,” inProc. IEEE ICASSP, 2024, pp. 11 831–11 835

    work page 2024

  5. [5]

    Tradition or inno- vation: A comparison of modern asr methods for forced alignment,

    R. Rousso, E. Cohen, J. Keshet, and E. Chodroff, “Tradition or inno- vation: A comparison of modern asr methods for forced alignment,” in Proc. Interspeech, 2024, pp. 1525–1529

    work page 2024

  6. [6]

    End-to-end speech recognition: A survey,

    R. Prabhavalkar, T. Hori, T. N. Sainath, R. Schl ¨uter, and S. Watanabe, “End-to-end speech recognition: A survey,”IEEE/ACM Trans. Audio, Speech, and Language Processing, vol. 32, pp. 325–351, 2023

    work page 2023

  7. [7]

    Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,

    A. Graves, S. Fern ´andez, F. Gomez, and J. Schmidhuber, “Connection- ist temporal classification: labelling unsegmented sequence data with recurrent neural networks,” inProceedings of the 23rd international conference on Machine learning. ACM, 2006, pp. 369–376

    work page 2006

  8. [8]

    Sequence Transduction with Recurrent Neural Networks

    A. Graves, “Sequence transduction with recurrent neural networks,” ArXiv:1211.3711, ICML Representation Learning Workshop, 2012

    work page Pith review arXiv 2012

Show all 50 references
  1. [9]

    Attention-based models for speech recognition,

    J. K. Chorowski, D. Bahdanau, D. Serdyuk, K. Cho, and Y . Bengio, “Attention-based models for speech recognition,” inNIPS, 2015, pp. 577–585

    2015

  2. [10]

    Listen, attend and spell: A neural network for large vocabulary conversational speech recognition,

    W. Chan, N. Jaitly, Q. V . Le, and O. Vinyals, “Listen, attend and spell: A neural network for large vocabulary conversational speech recognition,” inProc. IEEE ICASSP, 2016, pp. 4960–4964

    2016

  3. [11]

    Improved training of end-to- end attention models for speech recognition,

    A. Zeyer, K. Irie, R. Schl ¨uter, and H. Ney, “Improved training of end-to- end attention models for speech recognition,” inInterspeech, Hyderabad, India, Sep. 2018

    2018

  4. [12]

    SpeechGPT: Empowering large language models with intrinsic cross-modal conversational abilities,

    D. Zhang, S. Li, X. Zhang, J. Zhan, P. Wang, Y . Zhou, and X. Qiu, “SpeechGPT: Empowering large language models with intrinsic cross-modal conversational abilities,” inFindings of the Association for Computational Linguistics: EMNLP 2023, H. Bouamor, J. Pino, and K. Bali, Eds....

    2023

  5. [13]

    Qwen-Audio: Advancing universal audio understanding via unified large-scale audio-language models,

    Y . Chu, J. Xu, X. Zhou, Q. Yang, S. Zhang, Z. Yan, C. Zhou, and J. Zhou, “Qwen-Audio: Advancing universal audio understanding via unified large-scale audio-language models,” arXiv:2311.07919, 2023

    2023 arXiv

  6. [14]

    LLMs and speech: Integration vs. combination,

    R. Schmitt, A. Zeyer, M. Zeineldeen, R. Schl ¨uter, and H. Ney, “LLMs and speech: Integration vs. combination,” Arxiv:2603.15045, Mar. 2026. [Online]. Available: https://arxiv.org/abs/2603.15045

    2026 arXiv

  7. [15]

    Robust speech recognition via large-scale weak supervision,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. Mcleavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” inProc. ICML, ser. Proceedings of Machine Learning Research, A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scarlett, Eds....

    2023

  8. [16]

    WhisperX: Time-accurate speech transcription of long-form audio,

    M. Bain, J. Huh, T. Han, and A. Zisserman, “WhisperX: Time-accurate speech transcription of long-form audio,” inProc. Interspeech, 2023, pp. 4489–4493

    2023

  9. [17]

    CrisperWhisper: Accurate timestamps on verbatim speech transcriptions,

    M. Zusag, L. Wagner, and B. Thallinger, “CrisperWhisper: Accurate timestamps on verbatim speech transcriptions,” inProc. Interspeech, 2024, pp. 1265–1269

    2024

  10. [18]

    Whisper has an internal word aligner,

    S.-L. Yeh, Y . Meng, and H. Tang, “Whisper has an internal word aligner,” inProc. IEEE ASRU, 2025

    2025

  11. [19]

    Attention- constrained inference for robust decoder-only text-to-speech,

    H. Wang, C. Du, Y . Guo, S. Wang, X. Chen, and K. Yu, “Attention- constrained inference for robust decoder-only text-to-speech,” ArXiv 2404.19723, 2024. [Online]. Available: https://arxiv.org/abs/2404.19723

    2024 arXiv

  12. [20]

    DOA: Training-free decoder-only attention policy for long-form simultaneous translation with SpeechLLMs,

    S. Papi and L. Bentivogli, “DOA: Training-free decoder-only attention policy for long-form simultaneous translation with SpeechLLMs,” ArXiv 2605.31432, 2026. [Online]. Available: https://arxiv.org/abs/2605.31432

    2026 arXiv

  13. [22]

    Available: https://arxiv.org/abs/2601.18220

    [Online]. Available: https://arxiv.org/abs/2601.18220

  14. [23]

    Deep inside convolutional networks: Visualising image classification models and saliency maps,

    K. Simonyan, A. Vedaldi, and A. Zisserman, “Deep inside convolutional networks: Visualising image classification models and saliency maps,” ArXiv:1312.6034; ICLR Workshop Track, 2014

    2014 arXiv

  15. [24]

    The conformer encoder may reverse the time dimension,

    R. Schmitt, A. Zeyer, M. Zeineldeen, R. Schl ¨uter, and H. Ney, “The conformer encoder may reverse the time dimension,” inProc. IEEE ICASSP, Hyderabad, India, Apr. 2025, preprint ArXiv:2501.04521

    2025 arXiv

  16. [25]

    Saliency-driven word alignment interpretation for neural machine translation,

    S. Ding, H. Xu, and P. Koehn, “Saliency-driven word alignment interpretation for neural machine translation,” inProceedings of the Fourth Conference on Machine Translation (Volume 1: Research Papers), O. Bojar, R. Chatterjee, C. Federmann, M. Fishel, Y . Graham, B. Haddow, M. ...

    2019

  17. [26]

    Joint CTC/attention decoding for end-to-end speech recognition,

    T. Hori, S. Watanabe, and J. Hershey, “Joint CTC/attention decoding for end-to-end speech recognition,” inProc. ACL, R. Barzilay and M.-Y . Kan, Eds. Vancouver, Canada: Association for Computational Linguistics, Jul. 2017, pp. 518–529. [Online]. Available: https: //aclantholog...

    2017

  18. [27]

    Scaling speech tech- nology to 1,000+ languages,

    V . Pratap, A. Tjandra, B. Shi, P. Tomasello, A. Babu, S. Kundu, A. Elkahky, Z. Ni, A. Vyas, M. Fazel-Zarandi, A. Baevski, Y . Adi, X. Zhang, W.-N. Hsu, A. Conneau, and M. Auli, “Scaling speech tech- nology to 1,000+ languages,”Journal of Machine Learning Research, vol. 25, no...

    2024

  19. [28]

    TorchAudio 2.1: Advancing speech recognition, self-supervised learning, and audio processing components for PyTorch,

    J. Hwang, M. Hira, C. Chen, X. Zhang, Z. Ni, G. Sun, P. Ma, R. Huang, V . Pratap, Y . Zhang, A. Kumar, C.-Y . Yu, C. Zhu, C. Liu, J. Kahn, M. Ravanelli, P. Sun, S. Watanabe, Y . Shi, and Y . Tao, “TorchAudio 2.1: Advancing speech recognition, self-supervised learning, and audi...

    2023

  20. [29]

    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,” in Proc. NeurIPS, H. Larochelle, M. Ranzato, R. Hadsell, M. Balcan, and H. Lin, Eds., vol. 33. Curran Associates, Inc., 2020, pp. 12 449– 12 460. [...

    2020

  21. [30]

    Automatic phoneme recognition on TIMIT dataset with Wav2Vec 2.0,

    V . Phy, “Automatic phoneme recognition on TIMIT dataset with Wav2Vec 2.0,” 2022. [Online]. Available: https://huggingface.co/ vitouphy/wav2vec2-xls-r-300m-timit-phoneme

    2022

  22. [31]

    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,” inProc. Interspeech, 2022, pp. 2278–2282

    2022

  23. [32]

    Fast conformer with linearly scalable attention for efficient speech recognition,

    D. Rekesh, N. R. Koluguri, S. Kriman, S. Majumdar, V . Noroozi, H. Huang, O. Hrinchuk, K. Puvvada, A. Kumar, J. Balam, and B. Gins- burg, “Fast conformer with linearly scalable attention for efficient speech recognition,” inProc. IEEE ASRU, 2023, pp. 1–8

    2023

  24. [33]

    OWSM v4: Improving open whisper-style speech models via data scaling and cleaning,

    Y . Peng, S. Muhammad, Y . Sudo, W. Chen, J. Tian, C.-J. Lin, and S. Watanabe, “OWSM v4: Improving open whisper-style speech models via data scaling and cleaning,” inProc. Interspeech, 2025

    2025

  25. [34]

    OWSM-CTC: An open encoder-only speech foundation model for speech recognition, translation, and language identification,

    Y . Peng, Y . Sudo, M. Shakeel, and S. Watanabe, “OWSM-CTC: An open encoder-only speech foundation model for speech recognition, translation, and language identification,” inProc. ACL, 8 2024, pp. 10 192–10 209. [Online]. Available: https://aclanthology.org/2024. acl-long.549

    2024

  26. [35]

    Stateful conformer with cache-based inference for streaming automatic speech recognition,

    V . Noroozi, S. Majumdar, A. Kumar, J. Balam, and B. Ginsburg, “Stateful conformer with cache-based inference for streaming automatic speech recognition,” inProc. IEEE ICASSP, 2024, pp. 12 041–12 045

    2024

  27. [36]

    Efficient sequence transduction by jointly predicting tokens and durations,

    H. Xu, F. Jia, S. Majumdar, H. Huang, S. Watanabe, and B. Ginsburg, “Efficient sequence transduction by jointly predicting tokens and durations,” inProc. ICML, ser. Proceedings of Machine Learning Research, A. Krause, E. Brunskill, K. Cho, B. Engelhardt, S. Sabato, and J. Scar...

    2023

  28. [37]

    Emformer: Efficient memory transformer based acoustic model for low latency streaming speech recognition,

    Y . Shi, Y . Wang, C. Wu, C.-F. Yeh, J. Chan, F. Zhang, D. Le, and M. Seltzer, “Emformer: Efficient memory transformer based acoustic model for low latency streaming speech recognition,” inProc. IEEE ICASSP, 2021, pp. 6783–6787

    2021

  29. [38]

    TorchAudio: Building blocks for audio and speech processing,

    Y .-Y . Yang, M. Hira, Z. Ni, A. Astafurov, C. Chen, C. Puhrsch, D. Pollack, D. Genzel, D. Greenberg, E. Z. Yang, J. Lian, J. Hwang, J. Chen, P. Goldsborough, S. Narenthiran, S. Watanabe, S. Chintala, and V . Quenneville-B´elair, “TorchAudio: Building blocks for audio and spee...

    2022

  30. [39]

    OWLS: Scaling laws for multilingual speech recognition and translation models,

    W. Chen, J. Tian, Y . Peng, B. Yan, C.-H. H. Yang, and S. Watanabe, “OWLS: Scaling laws for multilingual speech recognition and translation models,” inProc. ICML, 2025. [Online]. Available: https://openreview.net/forum?id=xnPW7yYomF

    2025

  31. [40]

    V oxtral,

    Mistral AI, “V oxtral,” arXiv:2507.13264, 2025

    2025 arXiv

  32. [41]

    Phi-4-mini technical report: Compact yet powerful multimodal language models via mixture-of-loras,

    A. Abouelenin, A. Ashfaq, A. Atkinson, H. Awadalla, N. Bach, J. Bao, A. Benhaim, M. Cai, V . Chaudhary, C. Chen, D. Chen, D. Chen, J. Chen, W. Chen, Y .-C. Chen, Y . ling Chen, Q. Dai, X. Dai, R. Fan, M. Gao, M. Gao, A. Garg, A. Goswami, J. Hao, A. Hendy, Y . Hu, X. Jin, M. Kh...

    2025 arXiv

  33. [42]

    Less is more: Accurate speech recognition & translation without web-scale data,

    K. C. Puvvada, P. ˙Zelasko, H. Huang, O. Hrinchuk, N. R. Koluguri, K. Dhawan, S. Majumdar, E. Rastorgueva, Z. Chen, V . Lavrukhin, J. Balam, and B. Ginsburg, “Less is more: Accurate speech recognition & translation without web-scale data,” inProc. Interspeech, 2024, pp. 3964–3968

    2024

  34. [43]

    TIMIT acoustic-phonetic continuous speech corpus,

    J. S. Garofolo, L. F. Lamel, W. M. Fisher, J. G. Fiscus, D. S. Pallett, N. L. Dahlgren, and V . Zue, “TIMIT acoustic-phonetic continuous speech corpus,” Linguistic Data Consortium, Philadelphia, LDC93S1, 1993

    1993

  35. [44]

    The Buckeye corpus of conversational speech: Labeling conventions and a test of transcriber reliability,

    M. A. Pitt, K. Johnson, E. Hume, S. Kiesling, and W. Raymond, “The Buckeye corpus of conversational speech: Labeling conventions and a test of transcriber reliability,”Speech Communication, vol. 45, no. 1, pp. 89–95, 2005

    2005

  36. [45]

    Learning important features through propagating activation differences,

    A. Shrikumar, P. Greenside, and A. Kundaje, “Learning important features through propagating activation differences,” inProc. ICML, ser. Proceedings of Machine Learning Research, D. Precup and Y . W. Teh, Eds., vol. 70. PMLR, Aug. 2017, pp. 3145–3153. [Online]. Available: http...

    2017

  37. [46]

    Towards better understanding of gradient-based attribution methods for deep neural networks,

    M. Ancona, E. Ceolini, C. ¨Oztireli, and M. Gross, “Towards better understanding of gradient-based attribution methods for deep neural networks,” inProc. ICLR, 2018. [Online]. Available: https: //openreview.net/forum?id=Sy21R9JAW

    2018

  38. [47]

    Smooth- Grad: removing noise by adding noise,

    D. Smilkov, N. Thorat, B. Kim, F. Vi ´egas, and M. Wattenberg, “Smooth- Grad: removing noise by adding noise,” arXiv:1706.03825, 2017

    2017 arXiv

  39. [48]

    Sanity checks for saliency maps,

    J. Adebayo, J. Gilmer, M. Muelly, I. Goodfellow, M. Hardt, and B. Kim, “Sanity checks for saliency maps,” inProc. NeurIPS, S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa- Bianchi, and R. Garnett, Eds., vol. 31. Curran Associates, Inc.,

  40. [49]

    Available: https://proceedings.neurips.cc/paper files/ paper/2018/file/294a8ed24b1ad22ec2e7efea049b8737-Paper.pdf

    [Online]. Available: https://proceedings.neurips.cc/paper files/ paper/2018/file/294a8ed24b1ad22ec2e7efea049b8737-Paper.pdf

    2018

  41. [50]

    Axiomatic attribution for deep networks,

    M. Sundararajan, A. Taly, and Q. Yan, “Axiomatic attribution for deep networks,” inProc. ICML, ser. Proceedings of Machine Learning Research, D. Precup and Y . W. Teh, Eds., vol. 70. PMLR, Aug. 2017, pp. 3319–3328. [Online]. Available: https: //proceedings.mlr.press/v70/sundar...

    2017

  42. [51]

    Improving performance of deep learning models with axiomatic attribution priors and expected gradients,

    G. Erion, J. D. Janizek, P. Sturmfels, S. M. Lundberg, and S.-I. Lee, “Improving performance of deep learning models with axiomatic attribution priors and expected gradients,”Nature Machine Intelligence, vol. 3, no. 7, pp. 620–631, Jul. 2021

    2021

Pith tools

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