Zhizheng Wu武执政

News动态

Publications, awards, talks, releases, and group milestones.论文、获奖、报告、开源,以及课题组的大事小情。

2026

  • Keynote at ISCSLP 2026ISCSLP 2026 大会主旨报告

    Keynote talk “Controllable and Unified Modeling for Speech Generation”. The slides, with audio demos, are online.在 ISCSLP 2026 作大会主旨报告,题为“Controllable and Unified Modeling for Speech Generation”,谈语音生成的可控与统一建模。报告幻灯片(含音频样例)已上线。

  • Interview with Shenzhen Satellite TV深圳卫视专访

    An 18-minute interview for the programme 科创最前沿 on voice interaction, emotion recognition, and how we should live with AI (in Chinese).做客深圳卫视《科创最前沿》,18 分钟聊语音交互、情绪识别,以及人该怎样与 AI 相处。

  • The group at NCMMSC 2026课题组亮相 NCMMSC 2026

    Group members attended the 21st National Conference on Man-Machine Speech Communication in Hengqin, November 5–8.11 月 5–8 日,课题组师生赴横琴参加第二十一届全国人机语音通讯学术会议。

  • New students join the group新同学报到

    Five Ph.D. students and nine M.Phil. students started this term. See the group page.本学期迎来五位博士生和九位硕士生,详见团队页面。

  • TARS published at ACL 2026TARS 被 ACL 2026 主会录用

    Closing the Modality Reasoning Gap for Speech Large Language Models appears in the main conference.论文题为 Closing the Modality Reasoning Gap for Speech Large Language Models,关注语音大模型的模态推理鸿沟。

  • MimicLM published in Findings of ACL 2026MimicLM 被 ACL 2026 Findings 录用

    New work on zero-shot voice imitation from pseudo-parallel speech corpora.用伪平行语料做自回归建模,实现零样本声音模仿。

  • Three students graduate三位同学顺利毕业

    Xueyao Zhang and Yuancheng Wang completed their Ph.D. degrees, and Zihao Fang completed the M.Phil.Xueyao Zhang、Yuancheng Wang 获博士学位,Zihao Fang 获硕士学位。

  • SpeechJudge presented at ICLR 2026SpeechJudge 亮相 ICLR 2026

    A dataset, benchmark, and reward model centered on human judgments of speech naturalness.让机器像人一样判断语音自然度:一套数据集、评测基准和奖励模型。

  • FlexiCodec presented at ICLR 2026FlexiCodec 亮相 ICLR 2026

    A dynamic neural audio codec with controllable frame rates from 3 to 12.5 Hz.动态帧率的神经音频编解码器,帧率可在 3–12.5 Hz 之间自由调节。

  • VoxPrivacy and FlexiVoice presented at ICLR 2026VoxPrivacy、FlexiVoice 亮相 ICLR 2026

    A benchmark for interactional privacy in speech language models, and natural-language style control for zero-shot text-to-speech.前者评测语音大模型的交互隐私,后者用一句自然语言指令控制零样本合成的说话风格。

  • Keynote at the School of AI Open Day人工智能学院开放日主旨报告

    A talk on AI’s shift from information processing toward knowledge processing.主题:AI 正从信息处理走向知识处理。

  • GenSR-Pref published at AAAI 2026GenSR-Pref 被 AAAI 2026 录用

    Multi-metric preference alignment for generative speech restoration.用多指标偏好对齐提升生成式语音修复。

  • Four papers at Interspeech 2026Interspeech 2026 四篇论文

    WhispEar, Zero-VC, NV-Bench, and a study of entity binding failures in speech LLM reasoning.WhispEar、Zero-VC、NV-Bench,以及一项关于语音大模型推理中“实体绑定”失败的研究。

  • Two papers at ICASSP 2026ICASSP 2026 两篇论文

    NVSpeech, on modeling paralinguistic vocalizations, and SP-MCQA, on evaluating TTS intelligibility beyond the word level.NVSpeech 关注副语言发声建模,SP-MCQA 把合成语音的可懂度评测从词级往上推了一层。

  • Three articles in IEEE/ACM TASLPIEEE/ACM TASLP 三篇论文

    Vevo2, over-the-air adversarial attacks and detection for speaker verification, and text-aligned speech tokens.Vevo2、声纹识别的空中对抗攻击与检测,以及文本对齐的语音 token。

  • General Chair of IEEE CAI 2027将出任 IEEE CAI 2027 大会主席

    Zhizheng Wu will serve as General Chair of the IEEE Conference on Artificial Intelligence 2027.武执政将担任 IEEE Conference on Artificial Intelligence 2027 大会主席。

2025

  • Best Poster Award最佳海报奖

    Yuancheng Wang received the Best Poster Award at the 2025 声纹处理研究与应用学术研讨会.Yuancheng Wang 在 2025 声纹处理研究与应用学术研讨会上获最佳海报奖。

  • Invited talk at Huawei Media Tech Summit华为媒体技术峰会特邀报告

    Research progress in speech tokenization and neural audio codecs.分享语音 tokenization 与神经音频编解码的研究进展。

  • Invited talk at the RTE ConferenceRTE 大会特邀报告

    Research and application progress in speech processing technology.分享语音处理技术的研究与落地进展。

  • Honourable Mention and Best Paper Finalist两项论文荣誉

    TaDiCodec received an Honourable Mention at NYSF 2025, and Noro was named a Best Paper Finalist at APSIPA 2025.TaDiCodec 获 NYSF 2025 Honourable Mention,Noro 入围 APSIPA 2025 最佳论文。

  • Best Presentation Award at NCMMSC 2025NCMMSC 2025 最佳报告奖

    Congratulations to Jun’an Zhang.恭喜 Jun’an Zhang!

  • Papers accepted to NeurIPS 2025NeurIPS 2025 多篇论文录用

    Congratulations to Yuancheng Wang, Jiachen Zheng, Jun’an Zhang, Xueyao Zhang, Huan Liao, and collaborators.恭喜 Yuancheng Wang、Jiachen Zheng、Jun’an Zhang、Xueyao Zhang、Huan Liao 和各位合作者。

  • Paper accepted to IEEE/ACM TASLPIEEE/ACM TASLP 论文录用

    Congratulations to Haorui He, Chaoren Wang, and collaborators.恭喜 Haorui He、Chaoren Wang 和各位合作者。

  • Invited talk at Huawei Audio Tech Forum华为音频技术论坛特邀报告

    A talk on semantics-guided generative speech enhancement.主题:语义引导的生成式语音增强。

  • Leading CCF AATC Track 1牵头 CCF AATC 赛道一

    Jun’an Zhang led the Speech Restoration Challenge, attracting more than 100 teams.Jun’an Zhang 牵头组织语音修复挑战赛,吸引 100 多支队伍参赛。

  • Invited talk at IALP 2025IALP 2025 特邀报告

    Reinforcement learning for text-to-speech synthesis.主题:强化学习在语音合成中的应用。

  • Papers accepted to INTERSPEECH and DAFx 2025INTERSPEECH、DAFx 2025 论文录用

    New work from Yicheng Gu and collaborators.Yicheng Gu 与合作者的新工作。

  • Paper accepted to ACL 2025ACL 2025 论文录用

    Congratulations to Xueyao Zhang and Yuancheng Wang.恭喜 Xueyao Zhang 和 Yuancheng Wang。

  • Three papers accepted to ICLR 2025ICLR 2025 三篇论文录用

    Congratulations to Xueyao Zhang, Yuancheng Wang, Jun’an Zhang, and collaborators.恭喜 Xueyao Zhang、Yuancheng Wang、Jun’an Zhang 和各位合作者。

  • LOKI selected as an ICLR 2025 SpotlightLOKI 入选 ICLR 2025 Spotlight

    A synthetic data detection benchmark built with large multimodal models.用多模态大模型构建的合成数据检测基准。

2024

  • IEEE SLT 2024 in MacaoIEEE SLT 2024 在澳门举行

    Zhizheng Wu served as General Chair. The group presented five papers, including Amphion and Emilia, and received a Best Paper Finalist recognition.武执政担任大会主席;课题组发表五篇论文(含 Amphion、Emilia),并有论文入围最佳论文。

  • SD-Eval presented at NeurIPS 2024SD-Eval 亮相 NeurIPS 2024

    A benchmark dataset for spoken dialogue understanding beyond words.一个考察“听话听音”的评测集:理解口语对话中文字之外的信息。

  • Survey talk at Interspeech 2024Interspeech 2024 综述报告

    A survey talk on zero-shot text-to-speech synthesis.应邀作零样本语音合成综述报告。

  • Invited talk at CCF YOCSEF Hong KongCCF YOCSEF 香港论坛特邀报告

    Prospects of AI voice and audio generation applications, at the forum “Music Meets AI: A New Frontier?”在“Music Meets AI: A New Frontier?”论坛上分享 AI 语音与音频生成的应用前景。

  • NaturalSpeech 3 presented at ICML 2024NaturalSpeech 3 亮相 ICML 2024

    Zero-shot speech synthesis with factorized codec and diffusion models, in collaboration with Microsoft.与微软合作:用属性解耦的编解码器和扩散模型做零样本语音合成。

  • Associate Editor of IEEE Signal Processing Letters出任 IEEE Signal Processing Letters 副编辑

    Zhizheng Wu joined the editorial board.武执政加入该刊编委会。