In academy, I am currently a postdoctoral researcher at Tianjin University (TJU) since 2026, working with Prof. Tong Guo. Before that, I received my Ph.D. from Beijing University of Posts and Telecommunications in 2025, advised by Prof. Jingyu Wang. In 2023, I was a visiting student at ETH Zurich, working with Prof. Ce Zhang.

In industry, I am currently a research intern at Yijiahe’s LLM Lab, focusing on Auto Speech Recognition for our medical robotic system. I am also a research consultant at Chengmei Technology Co., Ltd, focusing on the joint optimization of intelligent sensors.

😄 Feel free to contact me: ningwanyi@126.com

Research Interests

• Auto Speech Recognition: End-to-end ASR, Target speaker extraction
• Distributed Machine Learning: Federated learning, Distributed optimization
• Efficient LLMs: Model compression, Parameter-efficient fine-tuning, Knowledge distillation

News!

• [2026-05] 🔥 New preprint: FormalASR: End-to-End Spoken Chinese to Formal Text. We present two compact end-to-end models (0.6B & 1.7B) that directly transcribe spoken Chinese into formal written text, achieving up to 37.4% relative CER reduction over verbatim baselines — no post-processing LLM needed at deployment time.

• [2026-02] I join the School of Precision Instrument and Optoelectronics Engineering, Tianjin University as a postdoctoral researcher.

• [2025-07] Our paper about federated LoRA fine-tuning in heterogeneous settings has been accepted by TNNLS.

• [2024-10] Our paper about lossless compression for fine-tuned foundation models has been accepted by NeurIPS.

• [2024-06] Our paper about federated knowledge distillation has been accepted by TSC.

Selected Publications

• [TNNLS’25] Wanyi Ning, Jingyu Wang, Qi Qi, Haifeng Sun, Daixuan Cheng, Cong Liu, Lei Zhang, Zirui Zhuang, Jianxin Liao. “Federated Fine-Tuning on Heterogeneous LoRAs With Error-Compensated Aggregation”. [Paper]

• [ACL’25] Minwei Zhang, Haifeng Sun, Jingyu Wang, Shaolong Li, Wanyi Ning, Qi Qi, Zirui Zhuang, Jianxin Liao. “ClusterAttn: KV Cache Compression under Intrinsic Attention Clustering”. [Paper]

• [NeurIPS’ 24] Wanyi Ning, Jingyu Wang, Qi Qi, Mengde Zhu, Haifeng Sun, Daixuan Cheng, Jianxin Liao, Ce Zhang. “Fm-delta: Lossless compression for storing massive fine-tuned foundation models”. [Paper].

• [Euro-Par’24] Mengde Zhu*, Wanyi Ning*, Qi Qi, Jingyu Wang, Zirui Zhuang, Haifeng Sun, Jun Huang, Jianxin Liao. “Fluk: protecting federated learning against malicious clients for internet of vehicles”. [Paper]

• [TSC’24] Wanyi Ning, Qi Qi, Jingyu Wang, Mengde Zhu, Shaolong Li, Guang Yang, Jianxin Liao. “One Teacher is Enough: A Server-Clueless Federated Learning With Knowledge Distillation”. [Paper]

• [ICML’23 workshop] Berivan Isik*, Hermann Kumbong*, Wanyi Ning*, Xiaozhe Yao*, Sanmi Koyejo, Ce Zhang. “Gpt-zip: Deep compression of finetuned large language models”. [Paper]

• [JSAC’21] Wanyi Ning, Haifeng Sun, Xiaoyuan Fu, Xiang Yang, Qi Qi, Jingyu Wang, Jianxin Liao, Zhu Han. “Following the correct direction: Renovating sparsified SGD towards global optimization in distributed edge learning”. [Paper]