WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917413719638016 |
|---|---|
| author | Chen, Yifu Ji, Shengpeng Chen, Qian Liang, Tianle Li, Yangzhuo Wang, Ziqing Wang, Wen Lu, Jingyu Wang, Haoxiao Pu, Xueyi Zhuo, Fan Zhao, Zhou |
| author_facet | Chen, Yifu Ji, Shengpeng Chen, Qian Liang, Tianle Li, Yangzhuo Wang, Ziqing Wang, Wen Lu, Jingyu Wang, Haoxiao Pu, Xueyi Zhuo, Fan Zhao, Zhou |
| contents | End-to-end spoken dialogue models have garnered significant attention because they offer a higher potential ceiling in expressiveness and perceptual ability than cascaded systems. However, the intelligence and expressiveness of current open-source spoken dialogue models often remain below expectations. Motivated by the success of online reinforcement learning(RL) in other domains, one might attempt to directly apply preference optimization to spoken dialogue models, yet this transfer is non-trivial. We analyze these obstacles from the perspectives of reward modeling and rollout sampling, focusing on how sparse preference supervision interacts with dense speech generation under shared-parameter updates. Based on the analysis, we propose a modality-aware adaptive post-training recipe that makes RL practical for spoken dialogue: it constrains preference updates to the semantic channel and improves acoustic behavior via explicit anchoring, while dynamically regulating their mixture from rollout statistics to avoid unreliable preference gradients. We evaluate the method across multiple spoken dialogue benchmarks and representative architectures, and observe consistent improvements in semantic quality and speech expressiveness. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_14932 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training Chen, Yifu Ji, Shengpeng Chen, Qian Liang, Tianle Li, Yangzhuo Wang, Ziqing Wang, Wen Lu, Jingyu Wang, Haoxiao Pu, Xueyi Zhuo, Fan Zhao, Zhou Artificial Intelligence End-to-end spoken dialogue models have garnered significant attention because they offer a higher potential ceiling in expressiveness and perceptual ability than cascaded systems. However, the intelligence and expressiveness of current open-source spoken dialogue models often remain below expectations. Motivated by the success of online reinforcement learning(RL) in other domains, one might attempt to directly apply preference optimization to spoken dialogue models, yet this transfer is non-trivial. We analyze these obstacles from the perspectives of reward modeling and rollout sampling, focusing on how sparse preference supervision interacts with dense speech generation under shared-parameter updates. Based on the analysis, we propose a modality-aware adaptive post-training recipe that makes RL practical for spoken dialogue: it constrains preference updates to the semantic channel and improves acoustic behavior via explicit anchoring, while dynamically regulating their mixture from rollout statistics to avoid unreliable preference gradients. We evaluate the method across multiple spoken dialogue benchmarks and representative architectures, and observe consistent improvements in semantic quality and speech expressiveness. |
| title | WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2604.14932 |