Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Group Relative Policy Optimization for Speech Recognition

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Reinforcement learning from text rewards cuts speech-recognition errors and hallucinations.

desk verdict A genuinely useful result on using GRPO to reduce hallucinations in LLM-based ASR, with a load-bearing but partly addressable missing-control issue. read the letter →

arxiv 2509.01939 v1 pith:DMMJWUEW submitted 2025-09-02 eess.AS

Prashanth Gurunath Shivakumar , Yile Gu , Ankur Gandhe , Ivan Bulyko This is my paper · ORCID
classification eess.AS
keywords speechrecognitionGRPOreinforcementlearningfromhumanfeedbackLLM-basedASRhallucinationreductiondomainadaptationworderrorraterewardrule-basedrewards
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 proposes adding a reinforcement-learning stage, Group Relative Policy Optimization (GRPO), after supervised fine-tuning of LLM-based speech recognition systems. Instead of training a separate reward model, the system scores its own generated transcripts with simple rule-based rewards such as word error rate, exact match, and total edit distance. The authors report relative WER reductions up to 18.4% on in-domain benchmarks, and much larger reductions on out-of-domain meeting speech where the baseline model hallucinates heavily. They also show that GRPO adaptation to a new domain beats conventional supervised fine-tuning on that domain while degrading less on other languages.

What carries the argument

Group Relative Policy Optimization (GRPO), a variant of PPO that omits the critic network and computes advantages from the mean and standard deviation of rewards across a group of sampled outputs. The policy objective is clipped like PPO's but uses group-relative advantages, plus an optional KL-divergence penalty to keep the policy near the reference model. In this paper, rewards are rule-based functions of the transcript: negated WER, negated total edit distance, or exact match against the reference text. The mechanism's work is to push the model toward outputs with lower edit distance to human transcripts, which the authors show translates into fewer insertions and substitutions on difficu

What would settle it

Run the same GRPO training recipe with multiple random seeds and compute confidence intervals on the held-out WER for Voxpopuli, MLS, and People's Speech. If the 2B model's small gains (e.g., MLS En 4.83 to 4.85 or the 0.1-point Voxpopuli changes) flip sign across seeds, the claim of consistent significant improvement would fail, while the large AMI effect would stand or fall on whether insertion reductions replicate across seeds.

Watch

Extended reading notes

Core claim

The paper claims that GRPO, a critic-free policy-gradient method originally developed for LLM reasoning, works as an effective post-training stage for LLM-based ASR. The key move is to define rewards directly from ground-truth transcripts, so no reward model is needed. Testing on 2B and 8B Llama-3-style models, the authors find that GRPO with a word-error-rate reward reduces insertions drastically on out-of-domain data, converting a model that produced 87.88% WER on AMI far-field speech into one at 37.59% WER while preserving or improving most in-domain results. They also find that GRPO with a small KL penalty performs comparably to unregularized variants, and that DAPO and Dr. GRPO modifica

Load-bearing premise

The reported improvements are computed from single evaluation runs without confidence intervals or significance tests, so the smaller gains could be evaluation noise rather than genuine model improvement.

Editorial extensions

If this is right

  • If the central claim holds, GRPO with rule-based rewards offers a simple, reward-model-free way to reduce ASR hallucinations, especially in far-field and meeting audio where insertion errors dominate.
  • The method provides a domain-adaptation recipe that improves target-domain WER while causing less regression on other languages than standard supervised fine-tuning.
  • The result suggests that reward design, not just architecture or data scale, is a practical lever for controlling LLM-based ASR behavior, since exact-match rewards rival WER rewards on clean data while WER rewards are safer on noisy data.
  • The reported reduction of 8B-model WER on AMI from above 100% to roughly 40% implies that RLHF can rescue large models that overfit in-domain patterns and hallucinate on unseen acoustics.
  • Because the loss is computed from sampled transcripts, the approach is compatible with any LLM-based ASR interface, so the same recipe could be layered onto other speech-LLM systems.

Reading between the lines

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

  • A natural extension would be semantic or task-level rewards, such as slot error rate or intent accuracy, which the paper itself mentions as future work; evidence here suggests edit-distance rewards may not align perfectly with downstream semantic goals.
  • The pronounced effect on AMI suggests GRPO is acting partly as a calibration step that suppresses the model's prior over common text patterns; one could test this hypothesis by measuring the per-token probability of insertions before and after GRPO.
  • The relative gains appear largest where baseline WER is high, implying that the benefit of RLHF may scale with domain mismatch; a testable prediction is that GRPO will show diminishing returns on already-clean, well-matched test sets.
  • Single-run evaluations without confidence intervals leave open the possibility that small reported gains are noise, so an independent multi-seed replication on the smaller benchmarks would clarify which of the improvements are real.
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

4 major / 4 minor

Summary. The paper proposes applying Group Relative Policy Optimization (GRPO) as an additional RL fine-tuning stage for decoder-only LLM-based automatic speech recognition. Starting from an SFT checkpoint, the policy is optimized by sampling G transcript hypotheses and using group-relative advantages with simple rule-based rewards: negative WER, exact match, and negative total edit distance, with a KL penalty to the reference. Experiments use 2B and 8B Llama3 models with a frozen conformer speech encoder, trained on roughly 88k hours of multilingual speech. The authors report WER improvements up to 18.4% relative, large reductions on out-of-domain AMI (e.g., 2B SFT AMI-SDM WER dropping from 87.88% to 37.59%), and favorable domain adaptation compared with continued SFT.

Significance. If substantiated, the result would be a useful first demonstration that RLHF-style policy optimization can improve LLM-based ASR, with practical benefits in insertion/hallucination reduction and out-of-domain robustness. The method is simple; the GRPO objective, advantage estimator, and reward definitions are standard and clearly presented. The rule-based rewards directly target the evaluation metrics, and the large AMI gains are unlikely to be pure noise. However, the manuscript does not yet rule out the alternative explanation that additional gradient updates, rather than the GRPO objective, account for the gains, and it lacks statistical validation for the smaller in-domain improvements. These issues are load-bearing for the central claims.

major comments (4)
  1. [§III-A, Tables II–III] The GRPO stage adds up to 5000 additional training steps on the same data used for SFT (fixed learning rate 1e-6, global batch size 64). No control continues SFT for the same number of steps, so the reported gains—especially the dramatic insertion-rate reductions on AMI—may reflect additional optimization rather than the GRPO objective. All ablations (β=0, DAPO, Dr. GRPO) retain an RL-style loss, so they do not isolate this confound. Please add a matched continued-SFT baseline (same data, learning rate, number of steps, checkpoint selection) for at least the 2B model on the Table II and Table III evaluations.
  2. [§IV, Tables II–III] All results appear to be single runs without confidence intervals or significance tests. Many differences are within a few tenths of a percent WER (e.g., 2B Voxpopuli En 7.92→7.82; MLS En 4.83→4.85 after GRPO(WER); 8B MLS overall 4.87→4.95). Such small differences can be evaluation noise. Report multiple decoding runs/seeds and significance tests, or at least WER confidence intervals, before claiming 'significant improvements' for the in-domain results. The large AMI gains are more robust but still need statistical quantification.
  3. [§IV-B, Table IV] The domain-adaptation comparison does not specify whether the SFT-adapted and GRPO-adapted models receive matched training budgets, learning rates, data, and checkpoint selection. If the SFT adaptation was run for fewer steps or with a different schedule, the comparison is not fair. Please state the exact setup for both adaptation conditions and, ideally, vary SFT steps to show the comparison is not an artifact of compute allocation.
  4. [Table III] The hallucination-reduction claim is supported only by insertion-rate changes. This is a reasonable proxy, but the manuscript should either define a direct hallucination metric or explicitly frame the claim as 'reduced insertion errors' rather than 'reduced hallucinations.' Also, the 8B SFT AMI WER values exceed 100% (100.26 and 227.98); this should be explained in the text so readers do not misread it as an error, and the corresponding insertion/deletion/substitution percentages should be interpreted carefully.
minor comments (4)
  1. [Eq. (3)] The notation in the min expression is ambiguous: clip is applied to the importance ratio πϑ, but the advantage A-hat is written outside the clip. Please write the standard form min(πϑ A-hat, clip(πϑ, 1−ε, 1+ε) A-hat) explicitly.
  2. [Eq. (5)] The KL expression as written is a pointwise Monte Carlo estimator of D_KL[πθ||πref], not the divergence itself. Please clarify that the expectation is over πθ, or write the divergence in closed form.
  3. [Introduction, §1] The claim that this is 'the first attempt at application of RLHF to LLMs to improve speech recognition' should be softened or qualified, given that Refs. [24]–[32] apply DPO/GRPO to speech understanding, audio QA, and TTS. If the claim is specifically 'first for ASR,' say so explicitly and distinguish from those prior works.
  4. [§IV, Role of RL Algorithms] The statement that 'both DAPO and Dr. GRPO outperform traditional GRPO in most cases' is not clearly supported by Table II (e.g., MLS En and Voxpopuli Es for DAPO are worse than GRPO(WER); several rows are mixed). Please quantify the number of languages/configurations that improve rather than making a blanket statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: GRPO directly optimizes explicit rule-based rewards, and all claimed improvements are evaluated on held-out or out-of-domain partitions.

full rationale

The paper's derivation chain is not circular. GRPO (Eq. 3-6) is an off-the-shelf policy-gradient objective combined with rule-based rewards (Eq. 7-9: WER, EM, ED). Optimizing WER on training utterances and then reporting WER on held-out partitions is direct optimization, not a hidden reduction: the evaluation sets are not used to fit any parameter, and the out-of-domain results (TEDLIUM, AMI in Table III) come from corpora excluded from both SFT and GRPO. The reported AMI gains are therefore not forced by construction. The only self-citation, [24] (ALign-SLM), appears in the related-work section as a prior DPO application to spoken language models and is not used to justify or derive the GRPO method, so it is not load-bearing. The absence of a continued-SFT control is a methodological confound that could explain some gains via additional gradient steps, but that is a correctness/attribution concern, not circularity: no equation or fitted parameter makes the reported improvements equal to the training objective by definition.

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

The central claim rests on the standard GRPO objective, a fixed pretrained encoder, and the assumption that single-run dev-set checkpoint selection yields reliable WER estimates. No new physical or algorithmic entities are introduced. Free parameters are training hyperparameters, not constants fitted to produce a derivation.

free parameters (4)
  • KL divergence coefficient beta = 0.04 (default), 0.0 (ablated)
    Controls policy deviation from the reference SFT model. The paper reports no significant WER divergence when beta=0 (Section IV).
  • Number of GRPO generations G = 6 and 10 tested
    Authors state the impact was insignificant and skip the results (Section IV, Number of Generations).
  • GRPO training hyperparameters = learning rate 1e-6, global batch 64, max 5000 steps
    Fixed choices described in Section III.A; no sensitivity analysis beyond the ablated beta and G.
  • Reward scaling = unspecified
    Mentioned as a varied configuration (Section III.A) but no values or results are reported.
assumptions (4)
  • domain assumption GRPO policy-gradient objective (Eq. 3-6) is an unbiased and stable update for LLM fine-tuning.
    Adopted from DeepSeekMath [20], DAPO [22], and Dr. GRPO [23]; the paper relies on the cited formulation, including the advantage estimator in Eq. 4, without re-deriving it.
  • domain assumption The frozen 2B-parameter Conformer encoder provides sufficient speech representations for the RL stage.
    Section III.A fixes the encoder during GRPO; if the frozen features are suboptimal for RL training, the reported gains may not transfer to other encoder/decoder combinations.
  • domain assumption The SFT checkpoint is a suitable reference policy for the KL penalty.
    Section III.A uses the fine-tuned model as reference. The comparison assumes this anchor is reasonable and that rule-based rewards do not induce reward hacking, which the paper explicitly claims to avoid.
  • domain assumption Checkpoint selection on the dev partition yields a representative model for test evaluation.
    Section III.B states best checkpoints are picked on the development partition; single-run test WERs are then assumed to be stable, which is standard but unvalidated here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Group Relative Policy Optimization for Speech Recognition." pith.science (2026). https://pith.science/paper/DMMJWUEW

@misc{pith2026250901939,
  author       = {Pith},
  title        = {Pith review of: Group Relative Policy Optimization for Speech Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DMMJWUEW}},
  note         = {Machine review of arXiv:2509.01939}
}
read the original abstract

Speech Recognition has seen a dramatic shift towards adopting Large Language Models (LLMs). This shift is partly driven by good scalability properties demonstrated by LLMs, ability to leverage large amounts of labelled, unlabelled speech and text data, streaming capabilities with auto-regressive framework and multi-tasking with instruction following characteristics of LLMs. However, simple next-token prediction objective, typically employed with LLMs, have certain limitations in performance and challenges with hallucinations. In this paper, we propose application of Group Relative Policy Optimization (GRPO) to enable reinforcement learning from human feedback for automatic speech recognition (ASR). We design simple rule based reward functions to guide the policy updates. We demonstrate significant improvements in word error rate (upto 18.4% relative), reduction in hallucinations, increased robustness on out-of-domain datasets and effectiveness in domain adaptation.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Reinforcement Learning for Data-Efficient Code-Switched ASR

    cs.CL 2026-07 conditional novelty 5.0 of 10

    RLVR with CER and script-fidelity rewards matches full-data LoRA SFT for code-switched ASR using only 10% TTS data across 10 language pairs, with zero-shot transfer to human speech.

Reference graph

Works this paper leans on

47 extracted references · 15 canonical work pages · cited by 1 Pith paper

  1. [1]

    On gener- ative spoken language modeling from raw audio,

    K. Lakhotia, E. Kharitonov, W.-N. Hsu, Y . Adi, A. Polyak, B. Bolte, T.-A. Nguyen, J. Copet, A. Baevski, A. Mohamed et al. , “On gener- ative spoken language modeling from raw audio,” Transactions of the Association for Computational Linguistics, vol. 9, pp. 1336–1354, 2021

    2021

  2. [2]

    V oxtlm: unified decoder-only models for consolidating speech recognition/synthesis and speech/text continuation tasks,

    S. Maiti, Y . Peng, S. Choi, J.-w. Jung, X. Chang, and S. Watan- abe, “V oxtlm: unified decoder-only models for consolidating speech recognition/synthesis and speech/text continuation tasks,” arXiv preprint arXiv:2309.07937, 2023

    arXiv 2023

  3. [3]

    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,” arXiv preprint arXiv:2305.11000, 2023

    arXiv 2023

  4. [4]

    Lauragpt: Listen, attend, understand, and regenerate audio with gpt,

    Z. Du, J. Wang, Q. Chen, Y . Chu, Z. Gao, Z. Li, K. Hu, X. Zhou, J. Xu, Z. Ma et al., “Lauragpt: Listen, attend, understand, and regenerate audio with gpt,” arXiv preprint arXiv:2310.04673 , 2023

    arXiv 2023

  5. [5]

    An embarrassingly simple approach for llm with strong asr capacity,

    Z. Ma, G. Yang, Y . Yang, Z. Gao, J. Wang, Z. Du, F. Yu, Q. Chen, S. Zheng, S. Zhang et al., “An embarrassingly simple approach for llm with strong asr capacity,” arXiv preprint arXiv:2402.08846 , 2024

    arXiv 2024

  6. [6]

    On decoder-only architecture for speech-to-text and large language model integration,

    J. Wu, Y . Gaur, Z. Chen, L. Zhou, Y . Zhu, T. Wang, J. Li, S. Liu, B. Ren, L. Liu et al., “On decoder-only architecture for speech-to-text and large language model integration,” in 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) . IEEE, 2023, pp. 1–8

    work page 2023

  7. [7]

    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 preprint arXiv:2311.07919 , 2023

    arXiv 2023

  8. [8]

    Salmonn: Towards generic hearing abilities for large language models,

    C. Tang, W. Yu, G. Sun, X. Chen, T. Tan, W. Li, L. Lu, Z. Ma, and C. Zhang, “Salmonn: Towards generic hearing abilities for large language models,” arXiv preprint arXiv:2310.13289 , 2023

    arXiv 2023

Show all 47 references
  1. [9]

    Connecting speech encoder and large language model for asr,

    W. Yu, C. Tang, G. Sun, X. Chen, T. Tan, W. Li, L. Lu, Z. Ma, and C. Zhang, “Connecting speech encoder and large language model for asr,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 12 637– 12 641

    2024

  2. [10]

    Prompting large language models with speech recognition abilities,

    Y . Fathullah, C. Wu, E. Lakomkin, J. Jia, Y . Shangguan, K. Li, J. Guo, W. Xiong, J. Mahadeokar, O. Kalinli et al., “Prompting large language models with speech recognition abilities,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processin...

    2024

  3. [11]

    Prompting large language models for zero-shot domain adaptation in speech recognition,

    Y . Li, Y . Wu, J. Li, and S. Liu, “Prompting large language models for zero-shot domain adaptation in speech recognition,” in 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) . IEEE, 2023, pp. 1–8

    2023

  4. [12]

    Lost in transcription, found in distribution shift: Demystifying hallucination in speech foundation models,

    H. Atwany, A. Waheed, R. Singh, M. Choudhury, and B. Raj, “Lost in transcription, found in distribution shift: Demystifying hallucination in speech foundation models,” arXiv preprint arXiv:2502.12414 , 2025

    2025 arXiv

  5. [13]

    Hallucinations in neural automatic speech recognition: Identifying errors and hallucinatory models,

    R. Frieske and B. E. Shi, “Hallucinations in neural automatic speech recognition: Identifying errors and hallucinatory models,” arXiv preprint arXiv:2401.01572, 2024

    2024 arXiv

  6. [14]

    Corrective and reinforcement learning for speaker-independent continuous speech recognition,

    K.-F. Lee and S. Mahajan, “Corrective and reinforcement learning for speaker-independent continuous speech recognition,” Computer Speech & Language, vol. 4, no. 3, pp. 231–245, 1990

    1990

  7. [15]

    Maximum entropy-based reinforcement learning using a confidence measure in speech recognition for telephone speech,

    C. Molina, N. B. Yoma, F. Huenup ´an, C. Garret ´on, and J. Wuth, “Maximum entropy-based reinforcement learning using a confidence measure in speech recognition for telephone speech,” IEEE Transactions on Audio, Speech, and Language Processing , vol. 18, no. 5, pp. 1041– 1052, 2009

    2009

  8. [16]

    End-to-end speech recognition sequence training with reinforcement learning,

    A. Tjandra, S. Sakti, and S. Nakamura, “End-to-end speech recognition sequence training with reinforcement learning,” IEEE Access, vol. 7, pp. 79 758–79 769, 2019

    2019

  9. [17]

    Reinforcement learning of speech recognition system based on policy gradient and hypothesis selection,

    T. Kala and T. Shinozaki, “Reinforcement learning of speech recognition system based on policy gradient and hypothesis selection,” in 2018 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2018, pp. 5759–5763

    2018

  10. [18]

    Leveraging modality-specific representations for audio-visual speech recognition via reinforcement learning,

    C. Chen, Y . Hu, Q. Zhang, H. Zou, B. Zhu, and E. S. Chng, “Leveraging modality-specific representations for audio-visual speech recognition via reinforcement learning,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 11, 2023, pp. 12 607–12 615

    2023

  11. [19]

    Prox- imal policy optimization algorithms,

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Prox- imal policy optimization algorithms,” arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  12. [20]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models,

    Z. Shao, P. Wang, Q. Zhu, R. Xu, J. Song, X. Bi, H. Zhang, M. Zhang, Y . Li, Y . Wuet al., “Deepseekmath: Pushing the limits of mathematical reasoning in open language models,” arXiv preprint arXiv:2402.03300 , 2024

    2024 arXiv

  13. [21]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,

    D. Guo, D. Yang, H. Zhang, J. Song, R. Zhang, R. Xu, Q. Zhu, S. Ma, P. Wang, X. Bi et al., “Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,” arXiv preprint arXiv:2501.12948 , 2025

    2025 arXiv

  14. [22]

    Dapo: An open-source llm reinforcement learning system at scale,

    Q. Yu, Z. Zhang, R. Zhu, Y . Yuan, X. Zuo, Y . Yue, T. Fan, G. Liu, L. Liu, X. Liu et al., “Dapo: An open-source llm reinforcement learning system at scale,” arXiv preprint arXiv:2503.14476 , 2025

    2025 arXiv

  15. [23]

    Understanding r1-zero-like training: A critical perspective,

    Z. Liu, C. Chen, W. Li, P. Qi, T. Pang, C. Du, W. S. Lee, and M. Lin, “Understanding r1-zero-like training: A critical perspective,” arXiv preprint arXiv:2503.20783 , 2025

    2025 arXiv

  16. [24]

    Align-slm: Textless spoken language models with rein- forcement learning from ai feedback,

    G.-T. Lin, P. G. Shivakumar, A. Gourav, Y . Gu, A. Gandhe, H.-y. Lee, and I. Bulyko, “Align-slm: Textless spoken language models with rein- forcement learning from ai feedback,” arXiv preprint arXiv:2411.01834, 2024

    2024 arXiv

  17. [25]

    Qwen2-audio technical report,

    Y . Chu, J. Xu, Q. Yang, H. Wei, X. Wei, Z. Guo, Y . Leng, Y . Lv, J. He, J. Lin et al. , “Qwen2-audio technical report,” arXiv preprint arXiv:2407.10759, 2024

    2024 arXiv

  18. [26]

    Qwen2. 5-omni technical report,

    J. Xu, Z. Guo, J. He, H. Hu, T. He, S. Bai, K. Chen, J. Wang, Y . Fan, K. Danget al., “Qwen2. 5-omni technical report,” arXiv preprint arXiv:2503.20215, 2025

    2025 arXiv

  19. [27]

    Omni-r1: Reinforcement learning for om- nimodal reasoning via two-system collaboration,

    H. Zhong, M. Zhu, Z. Du, Z. Huang, C. Zhao, M. Liu, W. Wang, H. Chen, and C. Shen, “Omni-r1: Reinforcement learning for om- nimodal reasoning via two-system collaboration,” arXiv preprint arXiv:2505.20256, 2025

    2025 arXiv

  20. [28]

    R1-omni: Explainable omni-multimodal emotion recognition with reinforcement learning,

    J. Zhao, X. Wei, and L. Bo, “R1-omni: Explainable omni-multimodal emotion recognition with reinforcement learning,” arXiv preprint arXiv:2503.05379, 2025

    2025 arXiv

  21. [29]

    Omni-r1: Do you really need audio to fine-tune your audio llm?

    A. Rouditchenko, S. Bhati, E. Araujo, S. Thomas, H. Kuehne, R. Feris, and J. Glass, “Omni-r1: Do you really need audio to fine-tune your audio llm?” arXiv preprint arXiv:2505.09439 , 2025

    2025

  22. [30]

    Sari: Structured audio reasoning via curriculum-guided reinforcement learning,

    C. Wen, T. Guo, S. Zhao, W. Zou, and X. Li, “Sari: Structured audio reasoning via curriculum-guided reinforcement learning,” arXiv preprint arXiv:2504.15900, 2025

    2025 arXiv

  23. [31]

    Reinforcement learning outperforms supervised fine-tuning: A case study on audio question answering,

    G. Li, J. Liu, H. Dinkel, Y . Niu, J. Zhang, and J. Luan, “Reinforcement learning outperforms supervised fine-tuning: A case study on audio question answering,” arXiv preprint arXiv:2503.11197 , 2025

    2025 arXiv

  24. [32]

    F5r-tts: Improving flow-matching based text-to-speech with group relative policy optimization,

    X. Sun, R. Xiao, J. Mo, B. Wu, Q. Yu, and B. Wang, “F5r-tts: Improving flow-matching based text-to-speech with group relative policy optimization,” arXiv preprint arXiv:2504.02407 , 2025

    2025 arXiv

  25. [33]

    Rein- forcement learning with a corrupted reward channel,

    T. Everitt, V . Krakovna, L. Orseau, M. Hutter, and S. Legg, “Rein- forcement learning with a corrupted reward channel,” arXiv preprint arXiv:1705.08417, 2017

    2017 arXiv

  26. [34]

    Fleurs: Few-shot learning evaluation of universal representations of speech,

    A. Conneau, M. Ma, S. Khanuja, Y . Zhang, V . Axelrod, S. Dalmia, J. Riesa, C. Rivera, and A. Bapna, “Fleurs: Few-shot learning evaluation of universal representations of speech,” in 2022 IEEE Spoken Language Technology Workshop (SLT). IEEE, 2023, pp. 798–805

    2022

  27. [35]

    MLS: A large-scale multilingual dataset for speech research,

    V . Pratap, Q. Xu, A. Sriram, G. Synnaeve, and R. Collobert, “MLS: A large-scale multilingual dataset for speech research,” arXiv preprint arXiv:2012.03411, 2020

    2012 arXiv

  28. [36]

    V oxpopuli: A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation,

    C. Wang, M. Riviere, A. Lee, A. Wu, C. Talnikar, D. Haziza, M. Williamson, J. Pino, and E. Dupoux, “V oxpopuli: A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation,” arXiv preprint arXiv:2101.00390 , 2021

    2021 arXiv

  29. [37]

    The people’s speech: A large-scale diverse english speech recognition dataset for commercial usage,

    D. Galvez, G. Diamos, J. Ciro, J. F. Cer ´on, K. Achorn, A. Gopi, D. Kanter, M. Lam, M. Mazumder, and V . J. Reddi, “The people’s speech: A large-scale diverse english speech recognition dataset for commercial usage,” arXiv preprint arXiv:2111.09344 , 2021

    2021 arXiv

  30. [38]

    Com- mon voice: A massively-multilingual speech corpus,

    R. Ardila, M. Branson, K. Davis, M. Henretty, M. Kohler, J. Meyer, R. Morais, L. Saunders, F. M. Tyers, and G. Weber, “Com- mon voice: A massively-multilingual speech corpus,” arXiv preprint arXiv:1912.06670, 2019

    1912 arXiv

  31. [39]

    Covost 2 and massively multilingual speech-to-text translation,

    C. Wang, A. Wu, and J. Pino, “Covost 2 and massively multilingual speech-to-text translation,” arXiv preprint arXiv:2007.10310 , 2020

    2007 arXiv

  32. [40]

    The llama 3 herd of models,

    A. Grattafiori, A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Vaughan et al. , “The llama 3 herd of models,” arXiv preprint arXiv:2407.21783 , 2024

    2024 arXiv

  33. [41]

    Durep: Dual-mode speech representation learning via asr-aware distillation,

    P. R. Male, S. N. Ray, H. Arsikere, A. Jaiswal, P. Swarup, P. Sen, D. Chakrabarty, K. V . V . Girish, N. Bhave, F. Weber, S. Bhattacharya, and S. Garimella, “Durep: Dual-mode speech representation learning via asr-aware distillation,” 2025. [Online]. Available: https://arxiv.o...

    2025 arXiv

  34. [42]

    Textually pretrained speech language models,

    M. Hassid, T. Remez, T. A. Nguyen, I. Gat, A. Conneau, F. Kreuk, J. Copet, A. Defossez, G. Synnaeve, E. Dupoux et al. , “Textually pretrained speech language models,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  35. [43]

    Spirit-lm: Interleaved spoken and written language model,

    T. A. Nguyen, B. Muller, B. Yu, M. R. Costa-Jussa, M. Elbayad, S. Popuri, C. Ropers, P.-A. Duquenne, R. Algayres, R. Mavlyutov et al. , “Spirit-lm: Interleaved spoken and written language model,” Transactions of the Association for Computational Linguistics , vol. 13, pp. 30–52, 2025

    2025

  36. [44]

    V oxtlm: Unified decoder-only models for consolidating speech recog- nition, synthesis and speech, text continuation tasks,

    S. Maiti, Y . Peng, S. Choi, J.-w. Jung, X. Chang, and S. Watanabe, “V oxtlm: Unified decoder-only models for consolidating speech recog- nition, synthesis and speech, text continuation tasks,” in ICASSP 2024- 2024 IEEE International Conference on Acoustics, Speech and Signal ...

    2024

  37. [45]

    RedPajama: an open dataset for train- ing large language models,

    T. Computer, “RedPajama: an open dataset for train- ing large language models,” 2023. [Online]. Available: https://github.com/togethercomputer/RedPajama-Data

    2023

  38. [46]

    The AMI meeting corpus: A pre-announcement,

    J. Carletta, S. Ashby, S. Bourban, M. Flynn, M. Guillemot, T. Hain, J. Kadlec, V . Karaiskos, W. Kraaij, M. Kronenthal et al. , “The AMI meeting corpus: A pre-announcement,” in International workshop on machine learning for multimodal interaction . Springer, 2005, pp. 28– 39

    2005

  39. [47]

    TED-LIUM 3: Twice as much data and corpus repartition for experiments on speaker adaptation,

    F. Hernandez, V . Nguyen, S. Ghannay, N. Tomashenko, and Y . Es- teve, “TED-LIUM 3: Twice as much data and corpus repartition for experiments on speaker adaptation,” in Speech and Computer: 20th International Conference, SPECOM 2018, Leipzig, Germany, September 18–22, 2018, Pr...

    2018

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.