Full-duplex speech-to-speech (S2S) models provide natural, low-latency conversational interaction and would benefit from the ability to use external tools an…
机构:NVIDIA
来源:arXiv 2609.19334 | AI4Papers 论文推荐平台