Large audio language models (LALMs) have achieved strong performance on standard speech and audio understanding tasks, yet their ability to systematically ge…
机构:上交
来源:arXiv 2610.34413 | AI4Papers 论文推荐平台