arXiv：cs.CL· Kshitij Ambilduke, Caroline Brun, Thibaut Thonet, Vassilina Nikoulina, Marcely Zanon Boito·· 13 小时前
语音 LLM 去刻板印象个性化：BASIS-Bench 如何评估模型何时该装聋作哑
Personalization Without Stereotyping: Do Speech LLMs Know When to Turn a Deaf Ear?
AI 导读
研究者提出 BASIS-Bench，从人口统计学鲁棒性、刻板偏见和个性化三个维度，区分语音 LLM 的有益个性化与刻板行为。对八个近期语音 LLM 的评测显示，多数模型虽能识别口音等人口特征，却无法将其转化为恰当的个性化回应；显式人口标签能提升个性化，但同时放大了文本骨干中已有的刻板关联。
正文
Abstract:Recent speech LLMs can leverage rich paralinguistic cues, enabling more complex and personalized interactions. However, little work investigates how this sensitivity can lead to stereotyping, and existing evaluations largely measure demographic disparities without distinguishing beneficial personalization from stereotypical behavior. We introduce BASIS-Bench, a benchmark designed to disentangle these two aspects through three axes: (1) demographic robustness, measuring disparities in content-based tasks across speakers and demographic classification capabilities; (2) stereotypical bias, measuring whether perceived demographic cues trigger unprompted stereotypical behavior; and (3) personalization, measuring whether models can tailor responses to demographic characteristics when explicitly requested. We evaluate eight recent speech LLMs and find that, despite often recognizing demographic cues, most fail to translate them into appropriate personalization. Explicit demographic labels improve personalization but also amplify stereotypical associations already present in the textual backbones. These findings highlight the need to train speech LLMs not only to perceive demographic cues, but to learn when they should shape their responses.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.13550 [cs.CL] |
| (or arXiv:2610.13550v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13550 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Marcely Zanon Boito [view email]
[v1]
Fri, 9 Oct 2026 13:06:22 UTC (942 KB)
来源：arXiv：cs.CL · arxiv.org