arXiv · Audio and Speech· Jiaqi Xiong, Yunjia Qi, Qi Cao, Yu Zheng, Yutong Zhang, Ziteng Wang, Ruofan Liao, Weisheng Xu, Sichen Liu·· 4 hr agoAI score46
DEAF:面向音频大语言模型声学忠实度的诊断评测基准
DEAF: A Benchmark for Diagnostic Evaluation of Acoustic Faithfulness in Audio Language Models
AI brief
研究者提出 DEAF 基准,包含超过 2,700 个冲突刺激,覆盖情感韵律、背景声音和说话人身份三个声学维度,用于诊断音频多模态大语言模型是否真正处理声学信号。
Source: arXiv · Audio and Speech · arxiv.org