Think, Verbalize, then Speak: Bridging Complex Thoughts and Comprehensible Speech
This addresses the mismatch between textual and verbal delivery in spoken dialogue systems, offering an incremental improvement for applications using LLMs in speech communication.
The paper tackles the problem of suboptimal spoken dialogue from LLMs by proposing a Think-Verbalize-Speak framework that decouples reasoning from speech delivery, enhancing speech naturalness and conciseness with minimal impact on reasoning across multiple benchmarks.
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