Saved in:
Bibliographic Details
Main Authors: Xie, Hao-Hui, Chung, Ho-Lam, Lin, Yi-Cheng, Lu, Ke-Han, Ren, Wenze, Chen, Xie, Lee, Hung-yi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.05094
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916008333148160
author Xie, Hao-Hui
Chung, Ho-Lam
Lin, Yi-Cheng
Lu, Ke-Han
Ren, Wenze
Chen, Xie
Lee, Hung-yi
author_facet Xie, Hao-Hui
Chung, Ho-Lam
Lin, Yi-Cheng
Lu, Ke-Han
Ren, Wenze
Chen, Xie
Lee, Hung-yi
contents Large Audio-Language Models (LALMs) typically struggle with localized dialectal prosody due to the scarcity of specialized corpora. We present TW-Sound580K, a Taiwanese audio-text instruction dataset developed through a Verify-Generate-Critique (VGC) protocol. This pipeline leverages Dual-ASR validation to filter 522K raw clips, subsequently expanding them into 580,000 high-fidelity instruction pairs using a teacher model. The dataset's utility is demonstrated through Tai-LALM, which fine-tunes a DeSTA 2.5-Audio-initialized backbone and incorporates a dynamic Dual-ASR Arbitration strategy to optimize transcription selection during inference. On the TAU Benchmark, Tai-LALM reaches 49.1% accuracy, marking a 6.5% absolute improvement over the zero-shot baseline (42.6% with ASR text conditioning). This confirms that integrating regional corpora with rigorous curation and dynamic arbitration significantly enhances LALM performance on localized speech.
format Preprint
id arxiv_https___arxiv_org_abs_2603_05094
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TW-Sound580K: A Regional Audio-Text Dataset with Verification-Guided Curation for Localized Audio-Language Modeling
Xie, Hao-Hui
Chung, Ho-Lam
Lin, Yi-Cheng
Lu, Ke-Han
Ren, Wenze
Chen, Xie
Lee, Hung-yi
Sound
Large Audio-Language Models (LALMs) typically struggle with localized dialectal prosody due to the scarcity of specialized corpora. We present TW-Sound580K, a Taiwanese audio-text instruction dataset developed through a Verify-Generate-Critique (VGC) protocol. This pipeline leverages Dual-ASR validation to filter 522K raw clips, subsequently expanding them into 580,000 high-fidelity instruction pairs using a teacher model. The dataset's utility is demonstrated through Tai-LALM, which fine-tunes a DeSTA 2.5-Audio-initialized backbone and incorporates a dynamic Dual-ASR Arbitration strategy to optimize transcription selection during inference. On the TAU Benchmark, Tai-LALM reaches 49.1% accuracy, marking a 6.5% absolute improvement over the zero-shot baseline (42.6% with ASR text conditioning). This confirms that integrating regional corpora with rigorous curation and dynamic arbitration significantly enhances LALM performance on localized speech.
title TW-Sound580K: A Regional Audio-Text Dataset with Verification-Guided Curation for Localized Audio-Language Modeling
topic Sound
url https://arxiv.org/abs/2603.05094