Shikang Zheng

CV | Google Scholar | Email | LinkedIn | WeChat

Hi! I am currently a third-year undergraduate student majoring in Artificial Intelligence at South China University of Technology. Previously, I was mainly focusing on efficient Diffusion Models at EPIC Lab of Shanghai Jiao Tong University under the guidance of Prof. Linfeng Zhang. I am currently working with Prof. Xihui Liu on training efficient World Action Models and accelerating Autoregressive Video Models. Starting in July, I will join Prof. Max Simchowitz's group onsite, focusing on real-time World Models.

My research interests mainly focus on Efficient Generative Models, World Models and VLMs. I am sincerely seeking PhD opportunities for Fall 2027, if you find our interest matches, please email me!

News

[Apr, 2026] Homepage launched.

Selected Publications

* denotes equal contribution; only first and co-first authored papers are listed

	Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers Shikang Zheng, Guantao Chen, Qinming Zhou, Yuqi Lin, Lixuan He, Chang Zou, Peiliang Cai, Jiacheng Liu, Linfeng Zhang International Conference on Learning Representations (ICLR), 2026 Oral Presentation
	A Flat Vocabulary or a Rich Hierarchy? Re-introducing Intrinsic Structure Transforms the Autoregressive Image Generation Lixuan He, Shikang Zheng International Conference on Machine Learning (ICML), 2026
	From Sketch to Fresco: Efficient Diffusion Transformer with Progressive Resolution Shikang Zheng, Guantao Chen, Lixuan He, Jiacheng Liu, Yuqi Lin, Chang Zou, Linfeng Zhang IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
	Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Diffusion Transformers Guantao Chen, Shikang Zheng, Yuqi Lin, Linfeng Zhang IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026
	Forecast then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers Shikang Zheng, Liang Feng, Xinyu Wang, Qinming Zhou, Peiliang Cai, Chang Zou, Jiacheng Liu, Yuqi Lin, Junjie Chen, Yue Ma, Linfeng Zhang AAAI Conference on Artificial Intelligence (AAAI), 2026
	Rethinking Token-wise Feature Caching: Accelerating Diffusion Transformers with Dual Feature Caching Chang Zou, Shikang Zheng, Evelyn Zhang, Runlin Guo, Haohang Xu, Zhengyi Shi, Conghui He, Xuming Hu, Linfeng Zhang IEEE Transactions on Image Processing (Volume: 35), 2026
	AViTS: Adaptive Spatiotemporal Token Selection for Efficient Dynamic-Resolution Generation Haoran Qin, Zhengan Yan, Shikang Zheng, Xiaobing Tu, Jiacheng Liu, Yuqi Lin, Chang Zou, JinShan Liu, Peiliang Cai, Xiantao Zhang, Jinkui Ren, Linfeng Zhang European Conference on Computer Vision (ECCV), 2026*
	Stellar White-light Emission Reveals Two Types of Flaring Light Curve Patterns Shikang Zheng, Suraj Sahu, Valery M. Nakariakov, Ding Yuan, Song Feng Philosophical Transactions of the Royal Society A

Preprints

	Dynamic Video Generation: Shaping Video Generation Across Time and Space Shikang Zheng, Jingkai Huang, Jiacheng Liu, Guantao Chen, Lixuan He, Yuqi Lin, Peiliang Cai, Linfeng Zhang
	Eyes in the Update? Learning Where Rewards Land in Multimodal Reasoning Lixuan He, Tianzhe Chu, Shulin Tian, Shikang Zheng, Feng Chen, Ziwei Liu, Xihui Liu, Yi Ma
	Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion Peiliang Cai, Evelyn Zhang, Jiacheng Liu, Hao Lin, Ruiqi Zhang, Weile Mo, Yue Ma, Shikang Zheng, Jiehang Huang, Dongrui Liu, Linfeng Zhang
	SpecEdit: Training-Free Acceleration for Diffusion based Image Editing via Semantic Locking Zhengan Yan, Shikang Zheng, Haoran Qin, Xiaobing Tu, Yinggui Wang, Jiacheng Liu, Jiaxuan Ren, Yuqi Lin, Peiliang Cai, Jinkui Ren, Xiantao Zhang, Linfeng Zhang

Research Experience

Education

University of California, Berkeley

Exchange Student in EECS (2025.1 - 2025.4)
Berkeley, CA

South China University of Technology

B.Eng. in Artificial Intelligence (2023.9 - 2027.7)
GPA: 3.77/4.00
Guangzhou, China

Talks

Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers

Oral presentation, ICLR 2026 main conference, Rio de Janeiro, Brazil

How to Accelerate Diffusion Models?

DLMath & Efficiency, Embedded AI Lab @TUG, Graz, Austria

Services

Conference reviewer: CVPR 2026, ICML 2026, ECCV 2026, NeurIPS 2026

Miscellanea

My Chinese name (surname first) is 郑世康.
I am from Shenzhen, Guangdong Province in China.

Shikang Zheng

News

Selected Publications

Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers

A Flat Vocabulary or a Rich Hierarchy? Re-introducing Intrinsic Structure Transforms the Autoregressive Image Generation

From Sketch to Fresco: Efficient Diffusion Transformer with Progressive Resolution

Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Diffusion Transformers

Forecast then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers

Rethinking Token-wise Feature Caching: Accelerating Diffusion Transformers with Dual Feature Caching

AViTS: Adaptive Spatiotemporal Token Selection for Efficient Dynamic-Resolution Generation

Stellar White-light Emission Reveals Two Types of Flaring Light Curve Patterns

Preprints

Dynamic Video Generation: Shaping Video Generation Across Time and Space

Eyes in the Update? Learning Where Rewards Land in Multimodal Reasoning

Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion

SpecEdit: Training-Free Acceleration for Diffusion based Image Editing via Semantic Locking

Research Experience

Princeton University, Computational Imaging Lab

Carnegie Mellon University, Machine Learning Department

The University of Hong Kong, MMLab

Shanghai Jiao Tong University, EPIC Lab

Education

University of California, Berkeley

South China University of Technology

Talks

Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers

How to Accelerate Diffusion Models?

Services

Miscellanea