arXiv：cs.CL· Yixian Tan, Aulia Adila, Jonathan Heng, Kunat Pipatanakul, Donghang Wu, Anocha Sutaveephamochanon·· 16 小时前
SEA-HELM Audio：评估多语言 ASR-LLM 级联管线在 SEA-HELM 各子集上的表现
SEA-HELM Audio: Evaluating Multilingual ASR-LLM Pipeline Performance Across SEA-HELM Subsets
AI 导读
研究者推出 SEA-HELM Audio，将文本基准 SEA-HELM 扩展至东南亚多语言语音评估，用合成语音对比中文、英语、马来语、印尼语和泰米尔语在语音与文本输入下的表现。在多项选择推理和指令遵循任务上，级联 ASR-LLM 管线出现明显精度下降：推理下降 9.00 至 18.50 个百分点，指令遵循下降 20.00 至 42.86 个百分点，其中泰米尔语降幅超过 40 个百分点。
正文
Abstract:We introduce SEA-HELM Audio, a Southeast Asian (SEA) multilingual evaluation framework that extends the text-based SEA-HELM benchmark to assess performance degradation in cascaded automatic speech recognition and large language model (ASR-LLM) pipelines. Using synthesized speech, we compare speech-input and text-input performance across five languages (Chinese, English, Malay, Indonesian, and Tamil) on multiple-choice reasoning and instruction-following tasks. We evaluate open-weight and commercial ASR-LLM systems using identical verifiers and quantify the speech gap as the difference between text and speech accuracy. Results show substantial degradation, with accuracy drops of 9.00 to 18.50 percentage points for multiple-choice reasoning and 20.00 to 42.86 points for instruction following. Tamil exhibits the largest degradation, exceeding 40 percentage points. Our findings highlight the limitations of text-only benchmarks in predicting voice-agent reliability and demonstrate the value of SEA-HELM Audio for evaluating multilingual voice-agent systems.
| Subjects: | Sound (cs.SD); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.19869 [cs.SD] |
| (or arXiv:2610.19869v1 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19869 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Anocha Sutaveephamochanon PhD [view email]
[v1]
Wed, 14 Oct 2026 09:29:36 UTC (39 KB)
来源：arXiv：cs.CL · arxiv.org