Pith. sign in

REVIEW 3 major objections 5 minor 38 references

X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A frame-synchronous dual-head model jointly predicts ASR tokens and turn states every 80 ms, outperforming streaming and cascaded baselines on bilingual turn-taking tests.

desk verdict A useful incremental architecture for frame-synchronous turn state prediction, but the headline accuracy comparison rests on an undocumented label mapping and one ASR claim contradicts its own table. read the letter →

arxiv 2608.10878 v1 pith:MZLG5FXN submitted 2026-08-11 cs.CL eess.AS

Kaiqi Fu , Rime Wen , Altman Lin , Shawn Qin , Roy Gan , Hao Wang , Qian Wang This is my paper · ORCID
classification cs.CLeess.AS
keywords turn-takingstreamingASRframe-synchronouspredictiondelayed-streammodelingdual-headarchitecturespokendialoguesystemsbackchanneldetectionEasyTurn
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 claims that turn-taking in spoken dialogue can be predicted frame by frame, at the same 80 ms cadence as streaming speech recognition, by adding a second prediction head to a delayed-stream ASR model. The two heads share the same streaming representations, so a single forward pass emits both transcript tokens and one of five turn states (<|idle|>, <|noidle|>, <|incomplete|>, <|complete|>, <|backchannel|>). Supervision comes from word-level turn labels projected onto the ASR token timeline, which aligns the two tasks at the frame level. On the bilingual EasyTurn test sets, the paper reports that this design beats the streaming baseline SoulX-Duplug on every reported accuracy metric while keeping comparable latency, and it matches or exceeds the best cascaded systems without their VAD segmentation delay. If correct, this removes the need for separate ASR and turn-detection modules in real-time dialogue systems.

What carries the argument

The central object is the dual-head delayed-stream decoder: a causal audio encoder, a temporal adapter, and a decoder-only language model that emits one token every 80 ms, with ASR and turn-state heads sharing the hidden state. The load-bearing mechanism is ASR-anchored supervision: word-level turn states are projected onto the same 80 ms positions as the word-boundary token [W] and its subword tokens, so both tasks optimize a frame-synchronous discrete timeline. The configurable delay $\tau$ is conditioned into the decoder through AdaRMSNorm, allowing one model to trade latency against accuracy across multiple operating points.

What would settle it

Publish the exact label-to-state mapping for EasyTurn and rescore X2-Turn by comparing each word's final non-idle prediction with the ground truth at word level; if the reported advantage over SoulX-Duplug disappears or drops below a few points under this scoring, the frame-synchronous advantage would be shown to come from utterance-level aggregation rather than from the proposed supervision.

Watch

Extended reading notes

Core claim

X2-Turn extends a pretrained delayed-stream ASR model (Voxtral Realtime) with a parallel turn-state head. At each 80 ms decoding step, the model's single hidden state feeds both the ASR vocabulary head and the turn-state head; the turn-state prediction never feeds back into decoding, so errors cannot corrupt transcription. The turn-state labels are constructed by assigning each word a semantic state with an LLM annotator, then placing the label at the same 80 ms positions as the word's boundary token and subword tokens. The paper's reported results include ACC_comp 91.00, ACC_incomp 93.00, and ACC_bc 96.00 on EasyTurn-zh at 288 ms latency, and ACC_comp 92.10 and ACC_incomp 84.60 on EasyTurn-en at 225 ms, all better than SoulX-Duplug on the corresponding metrics at similar or lower latency. The authors interpret this as showing that frame-synchronous joint modeling can outperform chunk- or utterance-level state prediction while retaining streaming ASR quality.

Load-bearing premise

The accuracy claims rest on an unspecified mapping between EasyTurn's original labels and the model's five turn states; if that mapping is looser than the states the model was trained on, the reported accuracies would not measure the intended construct.

Editorial extensions

If this is right

  • A single streaming model can serve both ASR and turn-state prediction; at inference no external ASR or VAD segmentation is needed, because both heads run on the same forward pass.
  • On EasyTurn-zh and EasyTurn-en, X2-Turn outperforms SoulX-Duplug on every reported metric (e.g., 91.00 vs 77.67 ACC_comp on Chinese) with similar latency (288 ms vs 295 ms).
  • Lowering $\tau$ gives a controlled latency-accuracy trade-off: on Chinese, reducing $\tau$ from 480 to 320 ms cuts latency from 288 to 120 ms while average accuracy changes from 92.00 to 90.67; on English, it cuts latency from 225 to 65 ms with average accuracy changing from 88.49 to 85.09.
  • Stage1-ASR at $\tau=480$ ms beats chunk-based streaming baselines on most test sets (e.g., AISHELL-1 1.48 vs 2.90 and 2.79), so frame-wise delayed-stream ASR remains competitive even before the turn head is added.
  • Since turn-state predictions are not fed back into decoding, turn-taking errors cannot degrade the ASR transcript.

Reading between the lines

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

  • Because the turn head and ASR head share hidden states and predictions are frame-aligned, the same architecture could be extended to jointly predict other frame-level dialogue events such as laughter or speaker change at the same 80 ms cadence; the paper does not test this.
  • The paper reports utterance-level accuracies, so a per-frame evaluation would sharpen the picture: if frame-level state transitions such as idle-to-incomplete are recovered accurately, the model could directly drive a TTS barge-in policy, an application the paper leaves implicit.
  • The $\tau$ sweep suggests a practical deployment recipe: pick $\tau=320$ ms when latency matters most, since accuracy degrades only mildly, a trade-off a system designer could exploit without retraining. This is our extrapolation from Table 2, not a paper claim.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes X2-Turn, a dual-head streaming model that extends the pretrained Voxtral Realtime delayed-stream ASR backbone with a parallel turn-state head. Both heads operate on shared causal decoder hidden states; word-level turn labels are generated by an LLM and projected onto the 80 ms frame positions of the ASR word-boundary tokens. Training uses two stages: a large-scale Chinese-English ASR adaptation stage and a joint ASR/turn-state fine-tuning stage on EasyTurn and Fisher subsets. Evaluation on the EasyTurn Chinese and English test sets reports turn-state accuracy and latency against cascaded and streaming baselines, together with ASR accuracy on standard benchmarks across different streaming delays. The central claim is that frame-synchronous turn-state prediction with ASR-anchored supervision yields accurate low-latency turn-taking without an auxiliary ASR model.

Significance. The architectural idea is timely and practical: integrating turn-state prediction into a streaming ASR decoder at frame level, with no auxiliary ASR at inference, addresses a real deployment bottleneck for spoken dialogue systems. The ASR-anchored supervision scheme is a clean way to align semantic turn labels with the audio frame timeline, and the controllable streaming delay τ provides a useful accuracy-latency trade-off. Strengths of the paper include the two-stage full fine-tuning on public corpora, bilingual experiments, the τ ablation, and the comparison against several recent baselines. If the evaluation is valid, the reported gains over SoulX-Duplug on streaming turn-state accuracy are a meaningful contribution. However, the quantitative claims are currently not fully substantiated because the mapping from EasyTurn's native labels to the paper's five-state taxonomy is unspecified, and because the ASR comparison table contains an unclear column structure that weakens the related claim.

major comments (3)
  1. [§2.3, §3.1, §4.1 (Table 1)] The paper never specifies how the EasyTurn test set's native turn-state labels are mapped to the proposed five-state taxonomy (<|idle|>, <|noidle|>, <|incomplete|>, <|complete|>, <|backchannel|>), nor how an utterance-level ground-truth label is formed for the 'last non-idle predicted state' comparison. EasyTurn is described in [8] as using four turn states, while the model is trained with LLM-generated word-level labels under a five-state scheme. Without an explicit, consistent mapping, the accuracy numbers in Table 1 may reflect the mapping choice rather than genuine turn-state quality. This is the primary load-bearing issue because the paper's central claim is accuracy superiority over SoulX-Duplug.
  2. [§4.3 (Table 3)] The claim that Stage2-Turn 'still matches or exceeds Freeze-Omni' is not supported by the table as printed. Table 3 lists four method columns (Uni-ASR, Freeze-Omni, Stage1-ASR, Stage2-Turn) but the rows contain five values, and the delay-header line offers only three delay labels. If the columns are read in the order given, the AISHELL-1 row shows a Stage2-Turn value of 3.94 against a Freeze-Omni value of 2.79, and the GigaSpeech row shows 12.55 against 12.23, both contradicting the claim. The column structure and the claim need to be corrected or qualified.
  3. [§4.1 (Table 1)] No error bars, confidence intervals, or significance tests are reported for the central accuracy comparisons. Some of the margins over SoulX-Duplug are modest, e.g., English ACC_comp is 92.10 vs 89.33 (a 2.77-point gap), and the number of test utterances per split is not given. Without this information, it is difficult to assess whether the reported differences are reliable, especially for the smaller English ACC_incomp gap of 5.27 points and the Chinese ACC_comp gap, which is the core evidence for the method's superiority.
minor comments (5)
  1. [§3.1] The LLM annotation procedure is described only through the phrase 'following the annotation criteria defined in Section 2.3', but Section 2.3 contains no prompt, no operational definition of 'complete' vs 'incomplete', and no consistency checks. The exact annotation instructions should be provided for reproducibility.
  2. [§4.1 (Table 1)] The caption should define what 'ground-truth utterance' means and how the utterance-level label is obtained from word-level labels. It should also clarify how backchannel predictions are treated on the English test set, which reportedly has no backchannel split.
  3. [§2.3 (Eq. 4)] Please specify the rounding convention in Eq. (4) (e.g., round-half-up vs floor) and clarify whether s_i is the forced-aligned word onset in seconds; the current notation leaves the units and the mapping to frame indices implicit.
  4. [§1 and throughout] The abbreviation 'V AD' should be written consistently as 'VAD', and 'V oxtral' should be 'Voxtral' in the running text.
  5. [§5 (Conclusion)] The final sentence is incomplete: 'Experiments on the bilingual EasyTurn test sets that X2-Turn achieves...' is missing a verb such as 'show' or 'demonstrate'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: X2-Turn is an empirical dual-head streaming model evaluated against an external benchmark; no prediction reduces to a fitted parameter or self-citation.

full rationale

X2-Turn is an empirical architecture paper rather than a derivation. The turn-state head is trained on word-level LLM annotations projected to ASR token frames (Eq. 4), and the headline accuracies in Table 1 are measured against the external EasyTurn Chinese/English test sets. The streaming delay τ is swept as an operating point (Table 2), not fitted to the test metric, and λ=0.1 is a fixed training weight. No parameter is fitted to a subset of the evaluation data and then renamed a prediction. The paper's references to Voxtral Realtime, EasyTurn, SoulX-Duplug, Qwen3, and other systems are independent prior work; there is no load-bearing self-citation. The main weakness—the paper never specifies how EasyTurn's native turn-state labels are mapped to the five-state taxonomy used for evaluation—is a construct-validity and reproducibility concern about what Table 1 actually measures, but it is not a circular reduction: the test labels are not defined in terms of the model's outputs or the paper's own training-label generation, and no equation in the paper makes the reported accuracy equal to the training objective by construction. Therefore no specific circular step can be exhibited, and the appropriate score is 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No new physical or conceptual entities are postulated beyond the turn-state vocabulary, which is an output label set rather than an invented entity. The free parameters are limited to the loss weight and the latency operating point, while the unstated label mapping and LLM annotation assumptions are the main epistemic burdens.

free parameters (2)
  • lambda (turn loss weight) = 0.1
    Chosen by hand to balance the ASR and turn-state objectives in the joint loss (Eq. 3); not fitted to test data.
  • tau (target streaming delay) = 480 ms for main results; 320, 400, 480 ms in ablations
    Configurable latency knob, sampled over 1 to 30 frames during training; the main comparison uses 480 ms, which gives the highest reported accuracy.
assumptions (3)
  • domain assumption Word-level turn state labels assigned by Qwen3.5-Plus are treated as ground truth for training.
    Section 2.3 and 3.1 use an LLM annotator to label each word's turn state; no quality assessment or human verification is reported.
  • domain assumption Qwen3-ForceAligner word boundary timestamps are accurate to within an 80 ms frame.
    Section 3.1: word onsets/offsets used for [W] placement and latency calculation come from this aligner; no alignment error analysis is given.
  • ad hoc to paper EasyTurn test set labels are comparable to the paper's five-state taxonomy without an explicit mapping.
    Section 4.1 compares accuracy against EasyTurn, but the mapping from the dataset's native labels (if different) to <|complete|>, <|incomplete|>, <|backchannel|> is not described.

how reviews work

0 comments
Cite this review

Pith. "Pith review of X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction." pith.science (2026). https://pith.science/paper/MZLG5FXN

@misc{pith2026260810878,
  author       = {Pith},
  title        = {Pith review of: X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MZLG5FXN}},
  note         = {Machine review of arXiv:2608.10878}
}
read the original abstract

Accurate and responsive turn-taking is essential for spoken dialogue systems, which must distinguish in real time between user interruptions, backchannels that should be ignored, and the completion of an utterance. Prior modular approaches typically optimize turn state prediction at the utterance or fixed-chunk level, creating a mismatch with the continuous turn state estimate, and often depend on an auxiliary ASR model, which limits responsiveness and increases overall system complexity. Therefore, we present X2-Turn, a frame-synchronous turn state prediction method via delayed-stream modeling. Specifically, building on the pretrained Voxtral Realtime model, we introduce a frame-synchronous turn state head that operates in parallel with the ASR head on shared streaming representations, jointly predicting ASR tokens and fine-grained turn states at the frame level. We evaluate our method on the bilingual Chinese-English Easy-Turn test sets, and the results demonstrate its effectiveness in achieving accurate turn-taking detection while maintaining low latency.

Figures

Figures reproduced from arXiv: 2608.10878 by the authors.

Figure 1
Figure 1. Overview of X2-Turn, the proposed frame-synchronous dual-head architecture, with a target delay of τ = 80 ms. Once the onset of a word has been observed and the target delay has elapsed, the ASR head emits a word-boundary token [W], while the turn state head simultaneously predicts the corresponding state. Subsequent subword tokens are then emitted frame by frame. 2. Method In this section, we first introduce the st… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 18 canonical work pages

  1. [8]

    Conclusion This paper presents X2-Turn, a frame-synchronous dual-head extension of a pretrained delayed-stream ASR model for joint streaming ASR and turn state prediction. A parallel turn state head shares causal decoder representations with the ASR head, with ASR-anchored supervision projecting word-level turn la- bels onto the native 80 ms token timelin...

    work page

  2. [1]

    To manage these complex conversational dynamics, a responsive system must continuously estimate fine-grained turn states

    Introduction Achieving natural spoken dialogue requires systems to seam- lessly handle continuous speech, backchannels, and user inter- ruptions while maintaining low latency [1]. To manage these complex conversational dynamics, a responsive system must continuously estimate fine-grained turn states. These states serve as the foundation for real-time dial...

    work page

  3. [2]

    We propose X2-Turn, which extends a pretrained delayed- stream ASR model with a parallel turn state head, en- abling joint frame-synchronous ASR and turn state prediction within a single streaming forward pass

    work page

  4. [3]

    We further introduce an ASR-anchored supervision method that projects word-level turn annotations onto the frame-level ASR token timeline

    We design a unified turn state label set that supports interrup- tion, turn completion, and backchannel detection. We further introduce an ASR-anchored supervision method that projects word-level turn annotations onto the frame-level ASR token timeline

    work page

  5. [4]

    X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction

    We conduct bilingual experiments on the EasyTurn Chinese and English test sets, validating the effectiveness of the pro- posed method for streaming turn state prediction under dif- ferent latency settings controlled byτ. arXiv:2608.10878v1 [cs.CL] 11 Aug 2026 Figure 1:Overview of X2-Turn, the proposed frame-synchronous dual-head architecture, with a targe...

    work page Pith review arXiv 2026

  6. [5]

    um”, “ah

    Method In this section, we first introduce the streaming ASR backbone and dual-head architecture, and then describe the turn state to- ken design and the construction of ASR-anchored turn state la- bels. An overview of the framework is shown in Fig. 1. 2.1. Dual-Head Modeling Our goal is to estimate turn states synchronously with ASR tran- scription as sp...

    work page

  7. [6]

    Data Preparation The corpora used in this work consist of two parts: Chinese- English ASR data and turn-taking data

    Experimental setup 3.1. Data Preparation The corpora used in this work consist of two parts: Chinese- English ASR data and turn-taking data. For the ASR data, we use AISHELL 1∼4 [17–20], AliMeeting [21], WenetSpeech [22], KeSpeech [23], LibriSpeech [24], Gi- gaSpeech [25], TED-LIUM [26], and V oxPopuli [27], totaling approximately 26k hours (14k hours in ...

    work page

  8. [7]

    Results and analysis We first compare the proposed method with cascaded baselines on the EasyTurn Chinese and English test sets. We then analyze the effect of the streaming delayτon turn-taking accuracy and latency, and finally compare ASR performance against chunk- based streaming baselines under different training stages. 4.1. Main results Table 1 compa...

    work page

Show all 38 references
  1. [9]

    Turn-taking in conversational systems and human- robot interaction: a review,

    G. Skantze, “Turn-taking in conversational systems and human- robot interaction: a review,”Computer Speech & Language, vol. 67, p. 101178, 2021

    2021

  2. [10]

    Generative spoken dialogue language modeling,

    T. A. Nguyen, E. Kharitonov, J. Copet, Y . Adi, W.-N. Hsu, A. Elkahky, P. Tomasello, R. Algayres, B. Sagot, A. Mohamed et al., “Generative spoken dialogue language modeling,”Trans- actions of the Association for Computational Linguistics, vol. 11, pp. 250–266, 2023

    2023

  3. [11]

    Moshi: a speech-text foundation model for real-time dialogue,

    A. D ´efossez, L. Mazar´e, M. Orsini, A. Royer, P. P´erez, H. J´egou, E. Grave, and N. Zeghidour, “Moshi: a speech-text foundation model for real-time dialogue,”arXiv preprint arXiv:2410.00037, 2024

    2024 arXiv

  4. [12]

    Freeze-omni: A smart and low latency speech- to-speech dialogue model with frozen LLM,

    X. Wang, Y . Li, C. Fu, Y . Zhang, Y . Shen, L. Xie, K. Li, X. Sun, and L. MA, “Freeze-omni: A smart and low latency speech- to-speech dialogue model with frozen LLM,” inForty-second International Conference on Machine Learning, 2025. [Online]. Available: https://openreview.ne...

    2025

  5. [13]

    Omniflatten: An end-to- end gpt model for seamless voice conversation,

    Q. Zhang, L. Cheng, C. Deng, Q. Chen, W. Wang, S. Zheng, J. Liu, H. Yu, C.-H. Tan, Z. Duet al., “Omniflatten: An end-to- end gpt model for seamless voice conversation,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pa...

    2025

  6. [14]

    Personaplex: V oice and role control for full duplex conversational speech models,

    R. Roy, J. Raiman, S.-g. Lee, T.-D. Ene, R. Kirby, S. Kim, J. Kim, and B. Catanzaro, “Personaplex: V oice and role control for full duplex conversational speech models,” inICASSP 2026- 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE...

    2026

  7. [15]

    Flexduo: A pluggable system for enabling full- duplex capabilities in speech dialogue systems,

    B. Liao, Y . Xu, J. Ou, K. Yang, W. Jian, P. Wan, and D. Zhang, “Flexduo: A pluggable system for enabling full- duplex capabilities in speech dialogue systems,”arXiv preprint arXiv:2502.13472, 2025

    2025 arXiv

  8. [16]

    Easy turn: Integrating acoustic and lin- guistic modalities for robust turn-taking in full-duplex spoken di- alogue systems,

    G. Li, C. Wang, H. Xue, S. Wang, D. Gao, Z. Zhang, Y . Lin, W. Li, L. Xiao, Z. Fuet al., “Easy turn: Integrating acoustic and lin- guistic modalities for robust turn-taking in full-duplex spoken di- alogue systems,” inICASSP 2026-2026 IEEE International Con- ference on Acousti...

    2026

  9. [17]

    Jal-turn: Joint acoustic- linguistic modeling for real-time and robust turn-taking detec- tion in full-duplex spoken dialogue systems,

    G. Yang, Y . Pan, S. Qiu, and N. Bai, “Jal-turn: Joint acoustic- linguistic modeling for real-time and robust turn-taking detec- tion in full-duplex spoken dialogue systems,”arXiv preprint arXiv:2603.26515, 2026

    2026

  10. [18]

    Fastturn: Unifying acoustic and streaming semantic cues for low-latency and robust turn detection,

    C. Wang, H. Xue, C. He, J. Hu, S. Wang, B. Wu, Y . Ji, J. Zheng, R. Chen, Z. Zhuet al., “Fastturn: Unifying acoustic and streaming semantic cues for low-latency and robust turn detection,”arXiv preprint arXiv:2604.01897, 2026

    2026 arXiv

  11. [19]

    Soulx-duplug: Plug-and-play streaming state prediction module for realtime full-duplex speech conversa- tion,

    R. Yan, W. Chen, Z. Liu, Z. Ma, H. Lin, H. Wen, H. Xie, J. Wu, Y . Liang, Y . Zhaoet al., “Soulx-duplug: Plug-and-play streaming state prediction module for realtime full-duplex speech conversa- tion,”arXiv preprint arXiv:2603.14877, 2026

    2026

  12. [20]

    JoyAI-Talker: Full-duplex speech interactive large model built for empathetic voice agents,

    JD.com, “JoyAI-Talker: Full-duplex speech interactive large model built for empathetic voice agents,”arXiv preprint arXiv:2608.01119, 2026

    2026 arXiv

  13. [21]

    Ten vad: A low-latency, lightweight and high- performance streaming voice activity detector (vad),

    T. Team, “Ten vad: A low-latency, lightweight and high- performance streaming voice activity detector (vad),” https://github.com/TEN-framework/ten-vad.git, 2025

    2025

  14. [22]

    Streaming sequence-to-sequence learning with de- layed streams modeling,

    N. Zeghidour, E. Kharitonov, M. Orsini, V . V olhejn, G. de Marmiesse, E. Grave, P. P ´erez, L. Mazar ´e, and A. D´efossez, “Streaming sequence-to-sequence learning with de- layed streams modeling,” Tech. Rep., 2025. [Online]. Available: https://arxiv.org/abs/2509.08753

    2025

  15. [23]

    V oxtral realtime,

    A. H. Liu, A. Ehrenberg, A. Lo, C.-Y . Sun, G. Lample, J.-M. Delignon, K. R. Chandu, P. von Platen, P. R. Mud- direddy, R. Aroraet al., “V oxtral realtime,”arXiv preprint arXiv:2602.11298, 2026

    2026 arXiv

  16. [24]

    Qwen3 technical report,

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lvet al., “Qwen3 technical report,”arXiv preprint arXiv:2505.09388, 2025

    2025 arXiv

  17. [25]

    Aishell-1: An open- source mandarin speech corpus and a speech recognition base- line,

    H. Bu, J. Du, X. Na, B. Wu, and H. Zheng, “Aishell-1: An open- source mandarin speech corpus and a speech recognition base- line,” in2017 20th conference of the oriental chapter of the inter- national coordinating committee on speech databases and speech I/O systems and assess...

    2017

  18. [26]

    Aishell-2: Transform- ing mandarin asr research into industrial scale,

    J. Du, X. Na, X. Liu, and H. Bu, “Aishell-2: Transform- ing mandarin asr research into industrial scale,”arXiv preprint arXiv:1808.10583, 2018

    2018 arXiv

  19. [27]

    Aishell-3: A multi- speaker mandarin tts corpus and the baselines,

    Y . Shi, H. Bu, X. Xu, S. Zhang, and M. Li, “Aishell-3: A multi- speaker mandarin tts corpus and the baselines,”arXiv preprint arXiv:2010.11567, 2020

    2010 arXiv

  20. [28]

    Aishell-4: An open source dataset for speech enhancement, separation, recognition and speaker diarization in conference scenario,

    Y . Fu, L. Cheng, S. Lv, Y . Jv, Y . Kong, Z. Chen, Y . Hu, L. Xie, J. Wu, H. Buet al., “Aishell-4: An open source dataset for speech enhancement, separation, recognition and speaker diarization in conference scenario,”arXiv preprint arXiv:2104.03603, 2021

    2021 arXiv

  21. [29]

    M2met: The icassp 2022 multi- channel multi-party meeting transcription challenge,

    F. Yu, S. Zhang, Y . Fu, L. Xie, S. Zheng, Z. Du, W. Huang, P. Guo, Z. Yan, B. Maet al., “M2met: The icassp 2022 multi- channel multi-party meeting transcription challenge,” inICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). I...

    2022

  22. [30]

    Wenetspeech: A 10000+ hours multi- domain mandarin corpus for speech recognition,

    B. Zhang, H. Lv, P. Guo, Q. Shao, C. Yang, L. Xie, X. Xu, H. Bu, X. Chen, C. Zenget al., “Wenetspeech: A 10000+ hours multi- domain mandarin corpus for speech recognition,” inICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IE...

    2022

  23. [31]

    Kespeech: An open source speech dataset of mandarin and its eight subdialects,

    Z. Tang, D. Wang, Y . Xu, J. Sun, X. Lei, S. Zhao, C. Wen, X. Tan, C. Xie, S. Zhouet al., “Kespeech: An open source speech dataset of mandarin and its eight subdialects,” inThirty-fifth conference on neural information processing systems datasets and bench- marks track (Round 2), 2021

    2021

  24. [32]

    Lib- rispeech: an asr corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “Lib- rispeech: an asr corpus based on public domain audio books,” in2015 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2015, pp. 5206–5210

    2015

  25. [33]

    Gigaspeech: An evolving, multi-domain asr corpus with 10,000 hours of transcribed audio,

    G. Chen, S. Chai, G. Wang, J. Du, W.-Q. Zhang, C. Weng, D. Su, D. Povey, J. Trmal, J. Zhanget al., “Gigaspeech: An evolving, multi-domain asr corpus with 10,000 hours of transcribed audio,” arXiv preprint arXiv:2106.06909, 2021

    2021 arXiv

  26. [34]

    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,” inInternational conference on speech and computer. Springer, 2018, pp. 198–208

    2018

  27. [35]

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

    C. Wang, M. Riviere, A. Lee, A. Wu, C. Talnikar, D. Haz- iza, M. Williamson, J. Pino, and E. Dupoux, “V oxpopuli: A large-scale multilingual speech corpus for representation learn- ing, semi-supervised learning and interpretation,” inProceedings of the 59th Annual Meeting of t...

    2021

  28. [36]

    The fisher corpus: A resource for the next generations of speech-to-text

    C. Cieri, D. Miller, and K. Walker, “The fisher corpus: A resource for the next generations of speech-to-text.” inLREC, vol. 4, 2004, pp. 69–71

    2004

  29. [37]

    Qwen3-asr technical report,

    X. Shi, X. Wang, Z. Guo, Y . Wang, P. Zhang, X. Zhang, Z. Guo, H. Hao, Y . Xi, B. Yanget al., “Qwen3-asr technical report,”arXiv preprint arXiv:2601.21337, 2026

    2026 arXiv

  30. [38]

    Uni-asr: Unified llm-based architecture for non-streaming and streaming automatic speech recognition,

    Y . Xia, J. Tang, J. Hou, G. Xu, and H. Yao, “Uni-asr: Unified llm-based architecture for non-streaming and streaming automatic speech recognition,”arXiv preprint arXiv:2603.11123, 2026

    2026

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.