Thinking While Speaking: Inference-Time Knowledge Transfer for Responsive and Intelligent Conversational Voice Agents

  • 类型：arxiv
  • 标识：2511.07397
  • 链接：https://arxiv.org/abs/2511.07397
  • 主分类：agent
  • 形态：method
  • 被引：0
  • 被引来源：Semantic Scholar + OpenAlex
  • S2被引：0
  • OpenAlex被引：0
  • 影响力被引：0
  • TLDR：Conversational infill is introduced, where a small talker model both immediately generates contextually grounded responses to hide the latency of an external reasoner model and fluently integrates streamed reasoner knowledge into its responses during inference.
  • OpenAlex ID：W4416159336
  • OpenAlex DOI：10.48550/arxiv.2511.07397
  • DOI：10.48550/arxiv.2511.07397
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://arxiv.org/pdf/2511.07397
  • OpenAlex更新：2026-07-19
  • 副分类：llm-infra
  • 待LLM分类：否
  • 标题中文：Thinking While Speaking: Inference-Time Knowledge Transfer for Responsive and Intelligent Conversational Voice Agents
  • TLDR中文：提出 conversational infill：让一个小型 talker 模型在外部 reasoner 模型产生结果前即时生成上下文相关的回复以掩盖延迟，并在推理过程中将 reasoner 流式输出的知识流畅地融合到回复中。
  • 来源文件：
  • /inbox/tom/_candidates/2026-06-29-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]