arXiv · cs.AI · Ruoyu He, Jisen Li, Anthony Ji, Changyi Yang, Zetian Li, Runduo Ma, Zhen Zhang, Zhongzhu Zhou · 10 小时前模型评分43
本地还是升级？实时双工语音模型的训练后专家路由
Local or Escalate? Post-Training Expert Routing for Real-Time Duplex Speech Models
导读
研究提出一种训练后路由方案：冻结 MiniCPM-o 4.5，仅用其语音输入输出，在原生“听转说”切换处的隐藏状态上拟合逻辑回归失败读出，判断当前问题该本地直答还是转交更强专家。
分类置信度
中 · 待人工复核
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “We study a post-training alternative.”
- “We freeze MiniCPM-o 4.5, using only its speech input and output, and fit a logistic failure readout on hidden states already available at its native listen-to-speak transition.”
- “The resulting gate requires no additional model generation and exposes thresholds that can be retuned without changing the duplex policy.”
来源：arXiv · cs.AI · arxiv.org