Guardado en:
Detalles Bibliográficos
Autores principales: Xia, Yinfeng, Tang, Jian, Hou, Junfeng, Xu, Gaopeng, Yao, Haitao
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2603.11123
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914386651643904
author Xia, Yinfeng
Tang, Jian
Hou, Junfeng
Xu, Gaopeng
Yao, Haitao
author_facet Xia, Yinfeng
Tang, Jian
Hou, Junfeng
Xu, Gaopeng
Yao, Haitao
contents Although the deep integration of the Automatic Speech Recognition (ASR) system with Large Language Models (LLMs) has significantly improved accuracy, the deployment of such systems in low-latency streaming scenarios remains challenging. In this paper, we propose Uni-ASR, a unified framework based on LLMs that integrates both non-streaming and streaming speech recognition capabilities. We propose a joint training paradigm that enables the system to seamlessly transition between two recognition modes without any architectural modifications. Furthermore, we introduce a context-aware training paradigm and a co-designed fallback decoding strategy, which can enhance streaming recognition accuracy without introducing additional latency. The experimental results demonstrate that Uni-ASR not only achieves competitive performance within non-streaming mode, but also demonstrates strong effectiveness in streaming scenarios under diverse latency constraints.
format Preprint
id arxiv_https___arxiv_org_abs_2603_11123
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Uni-ASR: Unified LLM-Based Architecture for Non-Streaming and Streaming Automatic Speech Recognition
Xia, Yinfeng
Tang, Jian
Hou, Junfeng
Xu, Gaopeng
Yao, Haitao
Sound
Computation and Language
Although the deep integration of the Automatic Speech Recognition (ASR) system with Large Language Models (LLMs) has significantly improved accuracy, the deployment of such systems in low-latency streaming scenarios remains challenging. In this paper, we propose Uni-ASR, a unified framework based on LLMs that integrates both non-streaming and streaming speech recognition capabilities. We propose a joint training paradigm that enables the system to seamlessly transition between two recognition modes without any architectural modifications. Furthermore, we introduce a context-aware training paradigm and a co-designed fallback decoding strategy, which can enhance streaming recognition accuracy without introducing additional latency. The experimental results demonstrate that Uni-ASR not only achieves competitive performance within non-streaming mode, but also demonstrates strong effectiveness in streaming scenarios under diverse latency constraints.
title Uni-ASR: Unified LLM-Based Architecture for Non-Streaming and Streaming Automatic Speech Recognition
topic Sound
Computation and Language
url https://arxiv.org/abs/2603.11123