大模型
arxiv.org ·
面向 Audio LLM 中英码切换语音识别的直接偏好优化
该论文聚焦 Audio LLM 在英语-普通话码切换语音转写中的系统性失败，归纳出语言遗漏、翻译替代转写和幻觉三种模式。作者用 DPO 构造偏好对（保留混合语言的回答为正例，模仿失败模式的为负例），在 100K 偏好对（570 小时）上训练三个 Audio LLM，观察到模型行为一致转向保留语言构成而非翻译。报告 MER 在分布内最多降低 89.6%、分布外降低 20.0%。论文已被 Interspeech 2026 收录。
AI推荐理由 arXiv 论文（已被 Interspeech 2026 收录），来源可信、事实明确。核心新信息：针对 Audio LLM 在中英码切换语音识别中的三类系统性失败（语言遗漏、翻译替代转写、幻觉），用 DPO 构造偏好对（chosen 保留混合语言、rejected 模仿失败模式），在 100K 对（570 小时）上训练三个 Audio LLM，报告 MER 在分布内最多下降 89.6%、分布外下降 20.0%。局限：摘要未给出模型名称、基线细节与评测集构成，89.6% 的降幅需结合具体基线理解，且为作者自报结果。对语音识别、多语言对齐与 DPO 应用有参考价值。
78
正文
Computer Science > Computation and Language
arXiv:2605.23975 (cs)
[Submitted on 13 May 2026 (v1), last revised 25 Sep 2026 (this version, v2)]
Title:Direct Preference Optimization for English-Mandarin Code-Switching Speech Recognition in Audio LLMs
Authors:Trung Nguyen Quang, Cheng Yi Lewis Won, Minh Duc Pham, Yingxu He, Shuo Sun, Ai Ti Aw
View a PDF of the paper titled Direct Preference Optimization for English-Mandarin Code-Switching Speech Recognition in Audio LLMs, by Trung Nguyen Quang and 5 other authors
View PDF
HTML (experimental)
Abstract:Audio large language models (Audio LLMs) exhibit systematic failures in transcribing code-switching speech despite strong multilingual capabilities. Focusing on English-Mandarin, we identify three failure modes: language omission, translation-instead-of-transcription, and hallucination. We apply Direct Preference Optimization (DPO) to align models, constructing preference pairs in which chosen responses preserve mixed-language content while rejected responses mimic failure patterns. Training three Audio LLMs on 100K pairs (570 hours), we observe consistent behavioral shifts: models learn to preserve language composition rather than translating when prompted for transcription. This alignment yields MER reductions up to 89.6% (in-distribution) and 20.0% (out-of-distribution). Our findings suggest DPO can effectively elicit correct code-switching transcription behavior from multilingual Audio LLMs.
Subjects:
Computation and Language (cs.CL); Sound (cs.SD)
Cite as:
arXiv:2605.23975 [cs.CL]
(or
arXiv:2605.23975v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2605.23975
Focus to learn more
arXiv-issued DOI via DataCite
Journal reference:
Proc. Interspeech 2026, pp. 6519-6524
Related DOI:
https://doi.org/10.21437/Interspeech.2026-110
Focus to learn more
DOI(s) linking to related resources
Submission history
From: Trung Nguyen Quang [view email]
[v1]
Wed, 13 May 2026 16:15:33 UTC (25 KB)
[v2]
Fri, 25 Sep 2026 15:13:17 UTC (25 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Direct Preference Optimization for English-Mandarin Code-Switching Speech Recognition in Audio LLMs, by Trung Nguyen Quang and 5 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CL
new
|
recent
| 2026-05
Change to browse by:
cs
cs.SD
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)