@inproceedings{sun-etal-2026-speech,
title = "Speech-{XL}: Towards Long-Form Speech Understanding in Large Speech Language Models",
author = "Sun, Haoqin and
Lyu, Chenyang and
Zhao, Shiwan and
Ni, Xuanfan and
Kong, Xiangyu and
Wang, Longyue and
Luo, Weihua and
Qin, Yong",
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.250/",
pages = "5255--5267",
ISBN = "979-8-89176-462-0",
abstract = "Despite the growing success of Large Speech Language Models (LSLMs) in processing short-term acoustic signals, their extension to long-form audio understanding is severely bottlenecked. This limitation stems from the limited context length and the exorbitant memory footprints required for long-form inference. In this work, we propose Speech-XL, a new model that capitalizes on the intrinsic key-value (KV) sparsification capacity of Large Language Models (LLMs) to achieve high-ratio speech input compression. Specifically, we introduce a novel special token, the Speech Summarization Token (SST), for each speech interval to encapsulate the intra-interval speech information into its associated KV pairs. The SST module is trained via instruction fine-tuning, employing a curriculum learning strategy where the SST learns to compress information in a progressive manner{---}advancing from low-ratio (simple) to high-ratio (challenging) compression. Despite using substantially less training data than large-scale audio foundation models, Speech-XL achieves competitive compression quality under the same backbone, maintaining a small gap to its uncompressed counterpart while substantially reducing long-form inference cost. By addressing the long-standing bottlenecks in long-form audio modeling, our approach offers a novel perspective on the condensation of extensive acoustic sequences."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sun-etal-2026-speech">
<titleInfo>
<title>Speech-XL: Towards Long-Form Speech Understanding in Large Speech Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Haoqin</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenyang</namePart>
<namePart type="family">Lyu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shiwan</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuanfan</namePart>
<namePart type="family">Ni</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiangyu</namePart>
<namePart type="family">Kong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Longyue</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weihua</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yong</namePart>
<namePart type="family">Qin</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>Despite the growing success of Large Speech Language Models (LSLMs) in processing short-term acoustic signals, their extension to long-form audio understanding is severely bottlenecked. This limitation stems from the limited context length and the exorbitant memory footprints required for long-form inference. In this work, we propose Speech-XL, a new model that capitalizes on the intrinsic key-value (KV) sparsification capacity of Large Language Models (LLMs) to achieve high-ratio speech input compression. Specifically, we introduce a novel special token, the Speech Summarization Token (SST), for each speech interval to encapsulate the intra-interval speech information into its associated KV pairs. The SST module is trained via instruction fine-tuning, employing a curriculum learning strategy where the SST learns to compress information in a progressive manner—advancing from low-ratio (simple) to high-ratio (challenging) compression. Despite using substantially less training data than large-scale audio foundation models, Speech-XL achieves competitive compression quality under the same backbone, maintaining a small gap to its uncompressed counterpart while substantially reducing long-form inference cost. By addressing the long-standing bottlenecks in long-form audio modeling, our approach offers a novel perspective on the condensation of extensive acoustic sequences.</abstract>
<identifier type="citekey">sun-etal-2026-speech</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.250/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>5255</start>
<end>5267</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Speech-XL: Towards Long-Form Speech Understanding in Large Speech Language Models
%A Sun, Haoqin
%A Lyu, Chenyang
%A Zhao, Shiwan
%A Ni, Xuanfan
%A Kong, Xiangyu
%A Wang, Longyue
%A Luo, Weihua
%A Qin, Yong
%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 sun-etal-2026-speech
%X Despite the growing success of Large Speech Language Models (LSLMs) in processing short-term acoustic signals, their extension to long-form audio understanding is severely bottlenecked. This limitation stems from the limited context length and the exorbitant memory footprints required for long-form inference. In this work, we propose Speech-XL, a new model that capitalizes on the intrinsic key-value (KV) sparsification capacity of Large Language Models (LLMs) to achieve high-ratio speech input compression. Specifically, we introduce a novel special token, the Speech Summarization Token (SST), for each speech interval to encapsulate the intra-interval speech information into its associated KV pairs. The SST module is trained via instruction fine-tuning, employing a curriculum learning strategy where the SST learns to compress information in a progressive manner—advancing from low-ratio (simple) to high-ratio (challenging) compression. Despite using substantially less training data than large-scale audio foundation models, Speech-XL achieves competitive compression quality under the same backbone, maintaining a small gap to its uncompressed counterpart while substantially reducing long-form inference cost. By addressing the long-standing bottlenecks in long-form audio modeling, our approach offers a novel perspective on the condensation of extensive acoustic sequences.
%U https://aclanthology.org/2026.findings-emnlp.250/
%P 5255-5267
Markdown (Informal)
[Speech-XL: Towards Long-Form Speech Understanding in Large Speech Language Models](https://aclanthology.org/2026.findings-emnlp.250/) (Sun et al., Findings 2026)
ACL
- Haoqin Sun, Chenyang Lyu, Shiwan Zhao, Xuanfan Ni, Xiangyu Kong, Longyue Wang, Weihua Luo, and Yong Qin. 2026. Speech-XL: Towards Long-Form Speech Understanding in Large Speech Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 5255–5267, Budapest, Hungary. Association for Computational Linguistics.