Large audio language models (LALMs) have made substantial progress in audio understanding. However, in high-throughput applications requiring frequent select…
机构:阿里
来源:arXiv 2610.21614 | AI4Papers 论文推荐平台