Ruixiang Mao

Ruixiang Mao

Master Student in Computer Science

I am a master's student in Computer Science at Northeastern University, working with the NEU NLP Lab.

My research interests include audio-language model understanding and reasoning, full-duplex speech interaction, MoE upcycling, automatic speech recognition, and speech translation.

I received my bachelor's degree in Information Management and Information Systems from Xiangtan University.

Email  /  Google Scholar  /  GitHub

Ruixiang Mao

News

Publications

Paper figure for When Scaling Fails

When Scaling Fails: Mitigating Audio Perception Decay of LALMs via Multi-Step Perception-Aware Reasoning

Ruixiang Mao, Xiangnan Ma, Dan Chen, Ziming Zhu, Yuan Ge, Aokai Hao, Haishu Zhao, Yifu Huo, Qing Yang, Kaiyan Chang, Xiaoqian Liu, Chenglong Wang, Qiaozhi He, Tong Xiao, Jingbo Zhu

EMNLP 2026 Main Conference

We study audio perception decay during long reasoning chains and propose a multi-step perception-aware reasoning strategy to keep audio-language models grounded in the input sound.

[paper] [code]

Paper figure for SPRI

SPRI: SVD-Partitioned Residual Initialization for Data-Constrained MoE Upcycling

Weiqiao Shan, Ruixiang Mao, Yuang Li, Yuhao Zhang, Yingfeng Luo, Tong Zheng, Chen Xu, Yucheng Qiao, Chunxiang Jin, Yi Yuan, Jingdong Chen, Tong Xiao, Jingbo Zhu

EMNLP 2026 Main Conference

We propose a structured initialization method for converting dense models into diverse sparse experts under limited supervised data.

[paper] [code]

Paper figure for M-CIF

M-CIF: Multi-Scale Alignment for CIF-Based Non-Autoregressive ASR

Ruixiang Mao, Xiangnan Ma, Qing Yang, Ziming Zhu, Yucheng Qiao, Yuan Ge, Tong Xiao, Shengxiang Gao, Zhengtao Yu, Jingbo Zhu

CCL 2026

We introduce multi-scale acoustic-text alignment to improve the robustness of non-autoregressive speech recognition across languages.

[paper] [code]

Education