Profile image

Kai Liu


second year PhD candidate
Department of Computer Science and Engineering, Shanghai Jiao Tong University
Office: Rm. 3-503, SEIEE Building
Email: kai_liu@sjtu.edu.cn

I am a second-year PhD student at the School of Computer Science (SCS), Shanghai Jiao Tong University (SJTU), advised by Prof. Yulun Zhang and Prof. Linghe Kong. Previously, I received my bachelor degree from SJTU in 2024.

My research interest is model compression and image quality assessment.

news

Oct 07, 2026 6 papers are accepted by NeurIPS 2026 🥳🥳🥳
May 01, 2026 Our papers CondiQuant and InfVSR are accepted by ICML 2026.
Oct 11, 2025 Our InfVSR and CLQ are released on arXiv.
Jul 09, 2025 Our survey on low-bit model quantization is released on arXiv.
May 01, 2025 Our paper BiMaCoSR is accepted by ICML 2025🥳🥳🥳

selected publications

  1. NeurIPS
    G \^ 2 TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models
    Junxian Li†, Kai Liu†, Zizhong Ding, Zhixin Wang, Zhikai Chen, Renjing Pei, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF Code
  2. NeurIPS
    TwinPrune: Density-Aware Two-Phase Token Pruning for Vision-Language Models
    Kai Liu†, Anqi Li†, Zhixin Wang, Zhikai Chen, Renjing Pei, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF
  3. NeurIPS
    PermuQuant: Lowering Per-Group Quantization Error by Reordering Channels for Diffusion Models
    Yongsen Cheng†, Kai Liu†, Kaiwen Tao, Junxian Li, Zhixin Wang, Zhikai Chen, Renjing Pei, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF Code
  4. NeurIPS
    QuantDemoire: Quantization with Outlier Aware for Image Demoir\backslash’eing
    Zheng Chen, Kewei Zhang, Xiaoyang Liu, Weihang Zhang, Mengfan Wang, Yifan Fu, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF Code
  5. NeurIPS
    DRScaffold: Boosting Dense-Scene Reasoning in Lightweight Vision Language Models
    Xinrui Shi†, Kai Liu†, Ziqing Zhang, Jianze Li, Anqi Li, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF Code
  6. NeurIPS
    Accelerating Rectified Flow Models via Trajectory-Aware Caching
    Xiao Liu†, Kai Liu†, Naiyang Guan, Hongliang Lu, Zhixin Wang, Zhikai Chen, Renjing Pei, and Yulun Zhang
    In Advances in Neural Information Processing Systems, 2026
    PDF
  7. WAIC
    Gate Entropy Reveals Quantization Fragility: Zero-Cost Mixed-Precision Allocation for Attention-Residual LLMs
    Kai Liu, Yulun Zhang*, and Linghe Kong*
    In WAIC Academic, 2026
  8. ICML
    InfVSR: Breaking Length Limits of Generic Video Super-Resolution
    Ziqing Zhang†, Kai Liu†, Zheng Chen, Xi Li, Yucong Chen, Bingnan Duan, Linghe Kong, and Yulun Zhang
    In International Conference on Machine Learning, 2026
    PDF Code
  9. arXiv
    CLQ: Cross-Layer Guided Orthogonal-based Quantization for Diffusion Transformers
    Kai Liu†, Shaoqiu Zhang†, Linghe Kong, and Yulun Zhang
    arXiv preprint arXiv:2509.24416, 2025
    Code
  10. arXiv
    Low-bit model quantization for deep neural networks: A survey
    Kai Liu†, Qian Zheng†, Kaiwen Tao†, Zhiteng Li, Haotong Qin, Wenbo Li, Yong Guo, Xianglong Liu, Linghe Kong*, Guihai Chen, Yulun Zhang*, and Xiaokang Yang
    arXiv preprint arXiv:2505.05530, 2025
    Code
  11. ICML
    CondiQuant: Condition Number Based Low-Bit Quantization for Image Super-Resolution
    Kai Liu†, Dehui Wang†, Zhiteng Li, Zheng Chen, Yong Guo, Wenbo Li, Linghe Kong*, and Yulun Zhang*
    In International Conference on Machine Learning, 2026
    PDF Code
  12. ICML
    BiMaCoSR: Binary One-Step Diffusion Model Leveraging Flexible Matrix Compression for Real Super-Resolution
    Kai Liu†, Kaicheng Yang†, Zheng Chen, Zhiteng Li, Yong Guo, Wenbo Li, Linghe Kong*, and Yulun Zhang*
    In International Conference on Machine Learning, 2025
    PDF Code
  13. arXiv
    Dog-IQA: Standard-guided Zero-shot MLLM for Mix-grained Image Quality Assessment
    Kai Liu†, Ziqing Zhang†, Wenbo Li, Renjing Pei, Fenglong Song, Xiaohong Liu, Linghe Kong*, and Yulun Zhang*
    arXiv, 2024
    PDF Code
  14. NeurIPS
    2DQuant: Low-bit Post-Training Quantization for Image Super-Resolution
    Kai Liu, Haotong Qin, Yong Guo, Xin Yuan, Linghe Kong*, Guihai Chen, and Yulun Zhang*
    In Advances in Neural Information Processing Systems, 2024
    PDF Code
  15. IEEE TMI
    Artifact Detection and Restoration in Histology Images With Stain-Style and Structural Preservation
    Jing Ke*, Kai Liu, Yuxiang Sun, Yuying Xue, Jiaxuan Huang, Yizhou Lu, Jun Dai, Yaobing Chen*, Xiaodan Han*, Yiqing Shen, and Dinggang Shen*
    In IEEE TMI, 2023
    PDF Code
  16. GLOBECOM
    Relayer-Enabled Sharding Blockchain for Satellite Internet with High Concurrency
    Hong Tang, Junqin Huang, Kai Liu, Linghe Kong, Guihai Chen, Shuangxi Cao, Chuyan Niu, and Yufeng Wei
    In GLOBECOM, 2023
    PDF
  17. CGI
    Large GAN Is All You Need
    Kai Liu*, Qingyang Wu, and Mengkun Xie
    In CGI, 2023
    PDF Code
  18. BIBMW
    HistDeblur: A Pyramid Trustworthy Framework for Blurry Histologic Artifacts Quality Control
    Yuxiang Sun†, Kai Liu†, Yiqing Shen, Xiaodan Han, and Jing Ke*
    In IEEE BIBM Workshop, 2023
    PDF Code