Decoder-only Conformer with Modality-aware Sparse Mixtures of Experts for ASR

• URL: http://arxiv.org/abs/2602.12546v1
• Date: Fri, 13 Feb 2026 02:53:54 GMT
• Title: Decoder-only Conformer with Modality-aware Sparse Mixtures of Experts for ASR
• Abstract summary: We present a decoder-only Conformer for automatic speech recognition (ASR) that processes speech and text in a single stack without external speech encoders or pretrained large language models (LLM)<n>The model uses a modality-aware sparse mixture of experts (MoE): disjoint expert pools for speech and text with hard routing and top-1 selection, embedded in hybrid-causality Conformer blocks.<n>On Common Voice 16.1 with a single multilingual model across five languages, our approach reduces average WER from 12.2% to 10.6%.
• Score: 9.626217175791572
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: We present a decoder-only Conformer for automatic speech recognition (ASR) that processes speech and text in a single stack without external speech encoders or pretrained large language models (LLM). The model uses a modality-aware sparse mixture of experts (MoE): disjoint expert pools for speech and text with hard routing and top-1 selection, embedded in hybrid-causality Conformer blocks (bidirectional for speech, causal for text). Training combines CTC on speech positions with label-smoothed cross-entropy for text generation. Our 113M-parameter model consistently improves WER over a 139M AED baseline on Librispeech (2.8% vs. 3.2% test-clean; 5.6% vs. 6.0% test-other). On Common Voice 16.1 with a single multilingual model across five languages, our approach reduces average WER from 12.2% to 10.6%. To our knowledge, this is the first randomly initialized decoder-only ASR that surpasses strong AED baselines via modality-aware routing and sparse MoE, achieving better accuracy with fewer active parameters and without alignment/adaptation modules.

Related papers

• Benchmarking Speech-to-Speech Translation Models [55.00303727199927]
  Speech-to-speech translation (S2ST) has advanced rapidly, but offline evaluation lacks a unified protocol.<n>We introduce a benchmarking framework integrating 46 metrics across eight dimensions.<n>We deploy it on 1,248 model-language configurations from FLEURS and CVSS.
  arXiv  Detail & Related papers  (2026-06-02T07:01:33Z)
• Qwen3-ASR Technical Report [71.87071808763484]
  We introduce Qwen3-ASR family, which includes two powerful all-in-one speech recognition models and a novel non-autoregressive speech forced alignment model.<n>Qwen3-ASR-1.7B and Qwen3-ASR-0.6B are ASR models that support language identification and ASR for 52 languages and dialects.
  arXiv  Detail & Related papers  (2026-01-29T06:58:13Z)
• Large-Scale Aspect-Based Sentiment Analysis with Reasoning-Infused LLMs [1.4732811715354455]
  Arctic-ABSA is a collection of powerful models for real-life aspect-based sentiment analysis (ABSA)<n>Our models are tailored to commercial needs, trained on a large corpus of public data alongside carefully generated synthetic data, resulting in a dataset 20 times larger than SemEval14.<n>A single multilingual model maintains 87-91% accuracy across six languages without degrading English performance.
  arXiv  Detail & Related papers  (2026-01-07T13:58:29Z)
• Exploring SSL Discrete Speech Features for Zipformer-based Contextual ASR [74.38242498079627]
  Self-supervised learning (SSL) based discrete speech representations are highly compact and domain adaptable.<n>In this paper, SSL discrete speech features extracted from WavLM models are used as additional cross-utterance acoustic context features in Zipformer-Transducer ASR systems.
  arXiv  Detail & Related papers  (2024-09-13T13:01:09Z)
• Cross-Speaker Encoding Network for Multi-Talker Speech Recognition [74.97576062152709]
  Cross-MixSpeaker. Network addresses limitations of SIMO models by aggregating cross-speaker representations. Network is integrated with SOT to leverage both the advantages of SIMO and SISO.
  arXiv  Detail & Related papers  (2024-01-08T16:37:45Z)
• From English to More Languages: Parameter-Efficient Model Reprogramming for Cross-Lingual Speech Recognition [50.93943755401025]
  We propose a new parameter-efficient learning framework based on neural model reprogramming for cross-lingual speech recognition. We design different auxiliary neural architectures focusing on learnable pre-trained feature enhancement. Our methods outperform existing ASR tuning architectures and their extension with self-supervised losses.
  arXiv  Detail & Related papers  (2023-01-19T02:37:56Z)
• Scaling Up Deliberation for Multilingual ASR [36.860327600638705]
  We investigate second-pass deliberation for multilingual speech recognition. Our proposed deliberation is multilingual, i.e., the text encoder encodes hypothesis text from multiple languages, and the decoder attends to multilingual text and audio. We show that deliberation improves the average WER on 9 languages by 4% relative compared to the single-pass model.
  arXiv  Detail & Related papers  (2022-10-11T21:07:00Z)
• Improving Deliberation by Text-Only and Semi-Supervised Training [42.942428288428836]
  We propose incorporating text-only and semi-supervised training into an attention-based deliberation model. We achieve 4%-12% WER reduction for various tasks compared to the baseline deliberation. We show that the deliberation model also achieves a positive human side-by-side evaluation.
  arXiv  Detail & Related papers  (2022-06-29T15:30:44Z)
• MixSpeech: Data Augmentation for Low-resource Automatic Speech Recognition [54.84624870942339]
  MixSpeech is a simple yet effective data augmentation method based on mixup for automatic speech recognition (ASR) We apply MixSpeech on two popular end-to-end speech recognition models including LAS (Listen, Attend and Spell) and Transformer. Experimental results show that MixSpeech achieves better accuracy than the baseline models without data augmentation.
  arXiv  Detail & Related papers  (2021-02-25T03:40:43Z)
• You Do Not Need More Data: Improving End-To-End Speech Recognition by Text-To-Speech Data Augmentation [59.31769998728787]
  We build our TTS system on an ASR training database and then extend the data with synthesized speech to train a recognition model. Our system establishes a competitive result for end-to-end ASR trained on LibriSpeech train-clean-100 set with WER 4.3% for test-clean and 13.5% for test-other.
  arXiv  Detail & Related papers  (2020-05-14T17:24:57Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.