About
I am a PhD student at OPTML Group at Michigan State University, advised by Prof. Sijia Liu. I received the MS degree in Computer Science at Northwestern University (NU) in June 2025, advised by Prof. Qi Zhu and Prof. Xiao Wang. Prior to NU, I obtained my B.E. in Tongji University in July 2023.
Away from research I am an avid astrophotographer. My work is collected in the photography gallery.
News
New preprint "Subspace Control: Turning Constrained Model Steering into Controllable Spectral Optimization" is available on arXiv: https://arxiv.org/abs/2604.04231
Our preprint "Forgetting to Forget: Attention Sink as A Gateway for Backdooring LLM Unlearning" is available on arXiv: https://arxiv.org/abs/2510.17021
Started my Ph.D. in Computer Science at Michigan State University, joining the OPTML Group advised by Prof. Sijia Liu.
Our paper about private downstream task adaptation of pre-trained transformers has been accepted to CVPR 2025.
Selected Publications
View All →* denotes equal contribution.
Subspace Control: Turning Constrained Model Steering into Controllable Spectral Optimization
Yancheng Huang, Changsheng Wang, Chongyu Fan, Yicheng Lang, Bingqi Shang, Yang Zhang, Mingyi Hong, Qing Qu, Alvaro Velasquez, Sijia Liu
arXiv preprint
Resolves objective-constraint interference in model steering by orthogonalizing the merged spectral subspace, then intervening only at the layers and steps that need it.
Forgetting to Forget: Attention Sink as A Gateway for Backdooring LLM Unlearning
Bingqi Shang*, Yiwei Chen*, Yihua Zhang, Bingquan Shen, Sijia Liu
arXiv preprint
Shows attention sinks act as gateways for backdooring LLM unlearning, so forgotten knowledge returns only when a hidden trigger is present.
Split Adaptation for Pre-trained Vision Transformers
Lixu Wang*, Bingqi Shang*, Yi Li, Payal Mohapatra, Wei Dong, Xiao Wang, Qi Zhu
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Splits a pre-trained ViT into quantized frontend and private backend so downstream adaptation protects both client data and model IP.