About me

Zhuokun Chen.

Making generative models
more efficient.

I am Zhuokun Chen, and I also go by Caesar. I received my master's degree from South China University of Technology in 2025, supervised by Mingkui Tan. In the same year, I began my PhD at Monash University, supervised by Bohan Zhuang and Jianfei Cai.

My research focuses on improving the efficiency of generative models, including image and video generation models and large language models (LLMs). I am particularly interested in sparse and linear attention, token reduction, and related techniques.

Contact: caesard216@gmail.com

Sparse / linear attentionToken reductionEfficient generative models
02 / Updates

News

  • Our History Forcing paper is now available on arXiv.

  • The HLA and HLA-WM projects have been released (first author).

  • The SAF project has been released (co-first author).

  • DASA has been accepted to NeurIPS 2026 (co-first author).

  • FlashBlock has been accepted to ICML 2026 (first author).

03 / Publications

Selected Publications

For more publications, see Google Scholar ↗.

* Equal contribution (co-first author).

04 / Academic service

Academic Service

Journal Reviewer

TPAMI · IJCV · TIP

Conference Reviewer

ICML 2026 · NeurIPS 2026 · ICLR 2026