Publications

Featured work

AAAI 2027 · Under review

AgenticASR: Refining Speech Recognition in Real-World Scenarios via an Agentic Approach

Zixuan Jiang*, Binghao Qiang*, Jiaying Chi*, Yanqiao Zhu, Kai Yu, Xie Chen†

• Defines Agentic Speech Recognition: turning spoken audio into clean text while preserving final intent.
• Separates ASR from a text Refiner, enabling bilingual online and offline cleanup.
• Releases AASR-Bench (6,637 fine-grained rubric items) and outperforms the other evaluated ASR families.

Featured work

TASLP · Under review

Towards Human-Like Interactive Speech Recognition With Agentic Correction and Semantic Evaluation

Zixuan Jiang*, Yanqiao Zhu*, Peng Wang*, Qinyuan Chen, Xinjian Zhao, Xipeng Qiu, Wupeng Wang, Zhifu Gao, Xiangang Li, Kai Yu, Xie Chen†

• Extends one-pass ASR into an interactive system with user feedback and semantic correction.
• Introduces an agent-based framework for interactive speech recognition.
• Develops the S²ER metric and ISS simulation framework for semantic evaluation.

Featured work

AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

Ziyang Ma, Zhikang Niu, Wenming Tu, Tianrui Wang, Ruiqi Yan, Junxi Liu, Yanru Huo, Nickk Huang, Yang Liu, Qicong Xie, Zeyu Xie, Hui Wang, Haitao Li, Zixuan Jiang, Yalin Li, Jie Fang, Yifan Duan, Zeyue Tian, Guangzheng Li, Haina Zhu, Shuyi Wang, Jinwen Wang, Mingyu Cui, Tian Tan, Auden, Sen Liang, Steve Yves, Shan Yang, Liefeng Bo, Zilong Zheng, Kai Yu, Eng-Siong Chng, Xie Chen

• Unifies speech generation and editing with natural-language instructions and reference audio.
• AuK-Flash: 4-step inference, approximately 4.5× faster under matched conditions.
• AuK-VAE: jointly trained on speech, music, and sound.

Featured work

Vol. 242 · pp. 860–872

DescribeEarth: Detailed localized captioning for remote sensing images

Kaiyu Li*, Zixuan Jiang*, Xiangyong Cao†, Jiayu Wang, Yuchen Xiao, Jing Yao, Chen Wu, Deyu Meng, Zhi Wang

• Defines a detailed localized captioning task for geospatial images.
• Builds the first remote-sensing model that can describe any specified region.
• Releases the associated dataset and evaluation benchmark.