WavReward: Spoken Dialogue Models With Generalist Reward Evaluators
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915507442024448 |
|---|---|
| author | Ji, Shengpeng Liang, Tianle Li, Yangzhuo Zuo, Jialong Fang, Minghui He, Jinzheng Chen, Yifu Liu, Zhengqing Jiang, Ziyue Cheng, Xize Zheng, Siqi Xu, Jin Lin, Junyang Zhao, Zhou |
| author_facet | Ji, Shengpeng Liang, Tianle Li, Yangzhuo Zuo, Jialong Fang, Minghui He, Jinzheng Chen, Yifu Liu, Zhengqing Jiang, Ziyue Cheng, Xize Zheng, Siqi Xu, Jin Lin, Junyang Zhao, Zhou |
| contents | End-to-end spoken dialogue models such as GPT-4o-audio have recently garnered significant attention in the speech domain. However, the evaluation of spoken dialogue models' conversational performance has largely been overlooked. This is primarily due to the intelligent chatbots convey a wealth of non-textual information which cannot be easily measured using text-based language models like ChatGPT. To address this gap, we propose WavReward, a reward feedback model based on audio language models that can evaluate both the IQ and EQ of spoken dialogue systems with speech input. Specifically, 1) based on audio language models, WavReward incorporates the deep reasoning process and the nonlinear reward mechanism for post-training. By utilizing multi-sample feedback via the reinforcement learning algorithm, we construct a specialized evaluator tailored to spoken dialogue models. 2) We introduce ChatReward-30K, a preference dataset used to train WavReward. ChatReward-30K includes both comprehension and generation aspects of spoken dialogue models. These scenarios span various tasks, such as text-based chats, nine acoustic attributes of instruction chats, and implicit chats. WavReward outperforms previous state-of-the-art evaluation models across multiple spoken dialogue scenarios, achieving a substantial improvement about Qwen2.5-Omni in objective accuracy from 53.4$\%$ to 91.5$\%$. In subjective A/B testing, WavReward also leads by a margin of 83$\%$. Comprehensive ablation studies confirm the necessity of each component of WavReward. All data and code will be publicly at https://github.com/jishengpeng/WavReward after the paper is accepted. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_09558 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | WavReward: Spoken Dialogue Models With Generalist Reward Evaluators Ji, Shengpeng Liang, Tianle Li, Yangzhuo Zuo, Jialong Fang, Minghui He, Jinzheng Chen, Yifu Liu, Zhengqing Jiang, Ziyue Cheng, Xize Zheng, Siqi Xu, Jin Lin, Junyang Zhao, Zhou Audio and Speech Processing Artificial Intelligence Machine Learning Multimedia Sound End-to-end spoken dialogue models such as GPT-4o-audio have recently garnered significant attention in the speech domain. However, the evaluation of spoken dialogue models' conversational performance has largely been overlooked. This is primarily due to the intelligent chatbots convey a wealth of non-textual information which cannot be easily measured using text-based language models like ChatGPT. To address this gap, we propose WavReward, a reward feedback model based on audio language models that can evaluate both the IQ and EQ of spoken dialogue systems with speech input. Specifically, 1) based on audio language models, WavReward incorporates the deep reasoning process and the nonlinear reward mechanism for post-training. By utilizing multi-sample feedback via the reinforcement learning algorithm, we construct a specialized evaluator tailored to spoken dialogue models. 2) We introduce ChatReward-30K, a preference dataset used to train WavReward. ChatReward-30K includes both comprehension and generation aspects of spoken dialogue models. These scenarios span various tasks, such as text-based chats, nine acoustic attributes of instruction chats, and implicit chats. WavReward outperforms previous state-of-the-art evaluation models across multiple spoken dialogue scenarios, achieving a substantial improvement about Qwen2.5-Omni in objective accuracy from 53.4$\%$ to 91.5$\%$. In subjective A/B testing, WavReward also leads by a margin of 83$\%$. Comprehensive ablation studies confirm the necessity of each component of WavReward. All data and code will be publicly at https://github.com/jishengpeng/WavReward after the paper is accepted. |
| title | WavReward: Spoken Dialogue Models With Generalist Reward Evaluators |
| topic | Audio and Speech Processing Artificial Intelligence Machine Learning Multimedia Sound |
| url | https://arxiv.org/abs/2505.09558 |