SoulX-Duplug-Eval

<div align="center"> <h1> SoulX-Duplug </h1> <p> Official code for enabling full-duplex speech interaction with<br> <b><em>SoulX-Duplug: Plug-and-Play Streaming State Prediction Module for Realtime Full-Duplex Speech Conversation</em></b> </p> <p> <img src="SoulX-Duplug-logo.png" alt="SoulX-Duplug Logo" style="width: 200px; height: 200px;"> </p> <p> </p> <a href="https://soulx-duplug.sjtuxlance.com/"><img src="https://img.shields.io/badge/🌐%20Online-Demo-blue" alt="Online Demo"></a> <a href="https://arxiv.org/abs/2603.14877"><img src="https://img.shields.io/badge/arXiv-2603.14877-B31B1B?logo=arxiv&logoColor=white.svg" alt="arXiv"></a> <a href="https://github.com/Soul-AILab/SoulX-Duplug"><img src='https://img.shields.io/badge/Github-Page-yellow?logo=github&logoColor=white.svg' alt="Github"></a> <a href="https://github.com/Soul-AILab/SoulX-Duplug"><img src="https://img.shields.io/badge/License-Apache%202.0-green.svg" alt="Apache-2.0"></a> </div> ## ✨ Overview SoulX-Duplug is a **plug-and-play streaming semantic VAD model** designed for real-time full-duplex speech conversation. Through text-guided streaming state prediction, SoulX-Duplug enables low-latency, semantic-aware streaming dialogue management. In addition to the core model, we also **open-source a [dialogue system](https://github.com/Soul-AILab/SoulX-Duplug/tree/dialogue-system) build on top of SoulX-Duplug**, which demonstrates the practicality of our model in real-world applications. To facilitate benchmarking and research in this area, we also release **[SoulX-Duplug-Eval](https://huggingface.co/datasets/Soul-AILab/SoulX-Duplug-Eval)**, a complementary evaluation set for benchmarking full-duplex spoken dialogue systems. ## 🧪 SoulX-Duplug-Eval - [Easy-Turn-Testset-en.zip](https://huggingface.co/datasets/Soul-AILab/SoulX-Duplug-Eval/blob/main/Easy-Turn-Testset-en.zip) contains the English counterpart of the original [Easy Turn testset](https://huggingface.co/datasets/ASLP-lab/Easy-Turn-Testset). It focuses on duplex state prediction and contains two categories: *Complete* and *Incomplete*. The *Complete* category consists of 318 semantically complete utterances. The *Incomplete* category contains 299 semantically incomplete samples. - [Full-Duplex-Bench-zh.zip](https://huggingface.co/datasets/Soul-AILab/SoulX-Duplug-Eval/blob/main/Full-Duplex-Bench-zh.zip) is a Chinese counterpart to [Full-Duplex-Bench](https://github.com/DanielLin94144/Full-Duplex-Bench). The dataset covers four representative interaction scenarios: *Turn Taking* (155 samples), *Pause Handling* (239 samples), *User Backchannel* (199 samples), and *User Interruption* (161 samples). The testsets follow the same format as Full-Duplex-Bench and can be directly evaluated using the official evaluation pipeline. ## 🔖 Citation If you find this work useful in your research, please consider citing: ```bibtex @misc{yan2026soulxduplug, title={SoulX-Duplug: Plug-and-Play Streaming State Prediction Module for Realtime Full-Duplex Speech Conversation}, author={Ruiqi Yan and Wenxi Chen and Zhanxun Liu and Ziyang Ma and Haopeng Lin and Hanlin Wen and Hanke Xie and Jun Wu and Yuzhe Liang and Yuxiang Zhao and Pengchao Feng and Jiale Qian and Hao Meng and Yuhang Dai and Shunshun Yin and Ming Tao and Lei Xie and Kai Yu and Xinsheng Wang and Xie Chen}, year={2026}, eprint={2603.14877}, archivePrefix={arXiv}, primaryClass={eess.AS}, url={https://arxiv.org/abs/2603.14877}, } ``` ## 📜 License This project is licensed under the [Apache 2.0 License](https://github.com/Soul-AILab/SoulX-Duplug/blob/main/LICENSE). ## 🙏 Acknowledgment We greatly thank [Easy Turn](https://github.com/ASLP-lab/Easy-Turn) and [Full-Duplex-Bench](https://github.com/DanielLin94144/Full-Duplex-Bench) for their contributions.