I am a second-year Ph.D. student in Artificial Intelligence at Shanghai Jiao Tong University, advised by Professor Yu Cheng. I received my B.S. in Computer Science and Technology from Nanjing University in 2024, where I graduated in the top 5% of my program.

My current research focuses on embodied intelligence, especially World Action Models. My previous research spans efficient generative AI, including diffusion model architectures, few-step generation, reinforcement learning, mixture-of-experts, and autoregressive video generation.

Google Scholar · GitHub · Download CV (Chinese)

News

• 2026 Jul: I joined Skywork’s Tiangong Robotics team as a research intern working on World Action Models.
• 2026 Jun: Flash-DMD was published at CVPR 2026.
• 2026 Feb: Less is More was published at AAAI 2026.
• 2025 Nov: I joined the Chinese University of Hong Kong as a Research Assistant.
• 2025 Sep: Skip-DiT was published at ICCV 2025.
• 2025 Jun: I joined Tencent Hunyuan as a research intern working on efficient visual generative models.
• 2024 Jul: MoE-RBench was published at ICML 2024.

Research Experience

• Skywork, Remain Dynamic, Research Intern, Jul. 2026 - Present
  World Action Models for embodied intelligence.
• The Chinese University of Hong Kong, Research Assistant, Nov. 2025 - Feb. 2026
  Elastic routing, sparsity enhancement, expert-parallel load balancing, and parameter pruning for mixture-of-experts models.
• Tencent, Hunyuan, Research Intern, Jun. 2025 - Nov. 2025
  Efficient image, video, and 3D generation through distillation, caching, and reinforcement learning; developed Flash-DMD.
• Shanghai AI Laboratory, Research Intern, Oct. 2023 - Jun. 2025
  Efficient DiT architectures, autoregressive video representation compression, and reliable mixture-of-experts models; developed Skip-DiT, VRC, and MoE-RBench.
• Nanjing University, Reasoning and Learning Lab, Research Intern, Jun. 2022 - Jun. 2023
  Semi-supervised and weakly supervised learning for medical image segmentation.

Selected Publications

1. Flash-DMD: Towards High-Fidelity Few-Step Image Generation with Efficient Distillation and Joint Reinforcement Learning
   Guanjie Chen, Shirui Huang, Kai Liu, Jianchen Zhu, Xiaoye Qu, Peng Chen, Yu Cheng, Yifu Sun.
   CVPR 2026. Paper

2. Skip-DiT: Towards Stabilized and Efficient Diffusion Transformers through Long-Skip-Connections with Spectral Constraints
   Guanjie Chen, Xinyu Zhao, Yucheng Zhou, Xiaoye Qu, Tianlong Chen, Yu Cheng.
   ICCV 2025. Paper · Code

3. MoE-RBench: Towards Building Reliable Language Models with Sparse Mixture-of-Experts
   Guanjie Chen, Xinyu Zhao, Tianlong Chen, Yu Cheng.
   ICML 2024. Paper