让大模型先思考再口头表达，提升对话自然度。
Think, Verbalize, then Speak: Bridging Complex Thoughts and Comprehensible Speech
- 分三步：思考→口头化→说话，分离推理与表达
- 口语化步骤使语音更自然简洁，推理准确率几乎不变
- 适合语音助手、智能客服等需流畅对话的场景
日益发展的语音对话系统越来越多地采用大语言模型（LLMs）以利用其强大的推理能力。然而，直接将LLM用于语音交流常因文本与口语表达之间的不匹配导致效果不佳。尽管现有方法尝试调整LLM输出以适应语音，但其对推理性能的影响仍缺乏深入研究。本文提出Think-Verbalize-Speak框架，将推理与口语表达解耦，保留LLM的完整推理能力。核心是引入‘口头化’这一中间步骤，将思维转化为自然、适合口语的文本。我们还提出了ReVerT——一种基于增量与异步摘要的低延迟口语化模块。在多个基准测试中，该方法显著提升了语音的自然度和简洁性，同时对推理性能影响极小。项目页面包含数据集与源码：https://yhytoto12.github.io/TVS-ReVerT。
原文摘要 · Abstract (English)
Spoken dialogue systems increasingly employ large language models (LLMs) to leverage their advanced reasoning capabilities. However, direct application of LLMs in spoken communication often yield suboptimal results due to mismatches between optimal textual and verbal delivery. While existing approaches adapt LLMs to produce speech-friendly outputs, their impact on reasoning performance remains underexplored. In this work, we propose Think-Verbalize-Speak, a framework that decouples reasoning from spoken delivery to preserve the full reasoning capacity of LLMs. Central to our method is verbalizing, an intermediate step that translates thoughts into natural, speech-ready text. We also introduce ReVerT, a latency-efficient verbalizer based on incremental and asynchronous summarization. Experiments across multiple benchmarks show that our method enhances speech naturalness and conciseness with minimal impact on reasoning. The project page with the dataset and the source code is available at https://yhytoto12.github.io/TVS-ReVerT
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。