Publications

* Equal Contribution, † Corresponding Author(s)

2025

  1. AAAI
    Numerical Pruning for Efficient Autoregressive Models
    Xuan Shen, Zhao Song, Yufa Zhou, Bo Chen, Jing Liu, Ruiyi Zhang, Ryan A Rossi, Hao Tan, Tong Yu, Xiang Chen, Yufan Zhou, Tong Sun, Pu Zhao, Yanzhi Wang, and Jiuxiang Gu
    The Association for the Advancement of Artificial Intelligence, 2025
  2. AAAI
    LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers
    Xuan Shen, Zhao Song, Yufa Zhou, Bo Chen, Yanyu Li, Yifan Gong, Kai Zhang, Hao Tan, Jason Kuen, Henghui Ding, Zhihao Shu, Wei Niu, Pu Zhao, Yanzhi Wang, and Jiuxiang Gu
    The Association for the Advancement of Artificial Intelligence, 2025
  3. NeurIPS
    Search for Efficient Large Language Models
    Xuan Shen, Pu Zhao, Yifan Gong, Zhenglun Kong, Zheng Zhan, Yushu Wu, Ming Lin, Chao Wu, Xue Lin, and Yanzhi Wang
    Conference on Neural Information Processing Systems, 2025

2024

  1. TCAD
    HotaQ: Hardware Oriented Token Adaptive Quantization for Large Language Models
    Xuan Shen, Zhaoyang Han, Lei Lu, Zhenglun Kong, Peiyan Dong, Zhengang Li, Yanyue Xie, Chao Wu, Miriam Leeser, Pu Zhao, Xue Lin, and Yanzhi Wang
    IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024
  2. AAAI
    Agile-Quant: Activation-Guided Quantization for Faster Inference of LLMs on the Edge
    Xuan Shen*, Peiyan Dong*, Lei Lu, Zhenglun Kong, Zhengang Li, Ming Lin, Chao Wu, and Yanzhi Wang
    The Association for the Advancement of Artificial Intelligence, 2024

2023

  1. CVPR
    DeepMAD: Mathematical Architecture Design for Deep Convolutional Neural Network
    Xuan Shen*, Yaohua Wang*, Ming Lin, Yilun Huang, Hao Tang, Xiuyu Sun, and Yanzhi Wang
    Conference on Computer Vision and Pattern Recognition, 2023
  2. IJCAI
    Data Level Lottery Ticket Hypothesis for Vision Transformers
    Xuan Shen, Zhenglun Kong, Minghai Qin, Peiyan Dong, Geng Yuan, Xin Meng, Hao Tang, Xiaolong Ma, and Yanzhi Wang
    International Joint Conference on Artificial Intelligence, 2023