@inproceedings{peng-etal-2026-tc,
title = "{TC}-{B}i{M}amba: Efficient Dynamic-Chunk Bidirectional Mamba for Unified Streaming and Offline {ASR}",
author = "Peng, Jing and
She, Qingshun and
Fang, Yangui and
Xi, Yu and
Wang, Shuai and
Yu, Kai",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.330/",
pages = "6976--6985",
ISBN = "979-8-89176-462-0",
abstract = "Unified streaming and offline ASR aims to deliver full-context accuracy and low-latency incremental decoding with a single model. Transformer/Conformer systems are effective, but long-form streaming inference still suffers from growing attention masks, historical caches, and memory states. Latency-Controlled BiMamba (LC-BiMamba) offers attention-free bidirectional modeling, yet its backward branch is confined to fixed chunks, tying the model to one latency. Extending it to dynamic chunk training expands mini-batches and slows training. We propose TC-BiMamba, whose Trans-Chunk formulation casts local-future backward modeling as sequence-level dynamic-chunk training over complete sequences. TC-BiMamba supports multiple chunk-wise latency settings, keeps the encoder/CTC path memory-stable on long-form inputs, and improves training scalability. On AISHELL-1, AISHELL-2, and LibriSpeech, it outperforms U2++ and remains competitive with externally reported LC-BiMamba results using fewer parameters, while achieving 1.3 times training speedup and 50{\%} GPU memory reduction."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="peng-etal-2026-tc">
<titleInfo>
<title>TC-BiMamba: Efficient Dynamic-Chunk Bidirectional Mamba for Unified Streaming and Offline ASR</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jing</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingshun</namePart>
<namePart type="family">She</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yangui</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Xi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuai</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Unified streaming and offline ASR aims to deliver full-context accuracy and low-latency incremental decoding with a single model. Transformer/Conformer systems are effective, but long-form streaming inference still suffers from growing attention masks, historical caches, and memory states. Latency-Controlled BiMamba (LC-BiMamba) offers attention-free bidirectional modeling, yet its backward branch is confined to fixed chunks, tying the model to one latency. Extending it to dynamic chunk training expands mini-batches and slows training. We propose TC-BiMamba, whose Trans-Chunk formulation casts local-future backward modeling as sequence-level dynamic-chunk training over complete sequences. TC-BiMamba supports multiple chunk-wise latency settings, keeps the encoder/CTC path memory-stable on long-form inputs, and improves training scalability. On AISHELL-1, AISHELL-2, and LibriSpeech, it outperforms U2++ and remains competitive with externally reported LC-BiMamba results using fewer parameters, while achieving 1.3 times training speedup and 50% GPU memory reduction.</abstract>
<identifier type="citekey">peng-etal-2026-tc</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.330/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>6976</start>
<end>6985</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T TC-BiMamba: Efficient Dynamic-Chunk Bidirectional Mamba for Unified Streaming and Offline ASR
%A Peng, Jing
%A She, Qingshun
%A Fang, Yangui
%A Xi, Yu
%A Wang, Shuai
%A Yu, Kai
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F peng-etal-2026-tc
%X Unified streaming and offline ASR aims to deliver full-context accuracy and low-latency incremental decoding with a single model. Transformer/Conformer systems are effective, but long-form streaming inference still suffers from growing attention masks, historical caches, and memory states. Latency-Controlled BiMamba (LC-BiMamba) offers attention-free bidirectional modeling, yet its backward branch is confined to fixed chunks, tying the model to one latency. Extending it to dynamic chunk training expands mini-batches and slows training. We propose TC-BiMamba, whose Trans-Chunk formulation casts local-future backward modeling as sequence-level dynamic-chunk training over complete sequences. TC-BiMamba supports multiple chunk-wise latency settings, keeps the encoder/CTC path memory-stable on long-form inputs, and improves training scalability. On AISHELL-1, AISHELL-2, and LibriSpeech, it outperforms U2++ and remains competitive with externally reported LC-BiMamba results using fewer parameters, while achieving 1.3 times training speedup and 50% GPU memory reduction.
%U https://aclanthology.org/2026.findings-emnlp.330/
%P 6976-6985
Markdown (Informal)
[TC-BiMamba: Efficient Dynamic-Chunk Bidirectional Mamba for Unified Streaming and Offline ASR](https://aclanthology.org/2026.findings-emnlp.330/) (Peng et al., Findings 2026)
ACL