Ph.D. Student @ HKUST(GZ)

Generative AI, efficient sampling, and visual foundation models.

I am Zikai Zhou, a Ph.D. researcher at HKUST(GZ). My work focuses on diffusion and flow-matching sampling, inference acceleration, video generation, and data-centric model improvement.

2026

Core contributor to Qwen Image 2.0, 3.0, and Turbo foundation model iterations.

ICML 2026

Gold Reviewer, Top 25%.

ICCV 2025

Golden Noise accepted at ICCV and cited 100+ times on Google Scholar.

Research

Efficient generation from algorithms to data flywheels.

Sampling and acceleration

Designing faster and more reliable sampling procedures for diffusion, flow matching, and masked generative transformers.

Data-centric generation

Building dataset condensation, filtering, distillation, and evaluation pipelines that close the loop between user signals and model training.

Video and image foundation models

Developing T2V, I2V, V2V, and text-to-image systems with stronger fidelity, controllability, and production readiness.

Evaluation pitfalls

Studying where generative evaluation breaks, including guidance, benchmark design, and automated pre-training feedback.

Publications

Research papers and technical reports

Full publication list
First page of Qwen-Image-2.0 Technical Report
Technical Report 2026Qwen Image

Qwen-Image-2.0 Technical Report

Qwen Team, including Zikai Zhou.

Documents the second-generation Qwen Image foundation model, including model iteration, data construction, training, and evaluation practices for large-scale text-to-image generation.

First page of Qwen-Image-VAE-2.0 Technical Report
Technical Report 2026VAE

Qwen-Image-VAE-2.0 Technical Report

Z. Zhang, D. Li, K. Cao, Y. Wu, C. Wu, Y. Wu, L. Peng, H. Meng, J. Li, J. Zhang, et al.

Presents the visual autoencoding component behind Qwen Image, focusing on compact visual representation, reconstruction quality, and downstream generation fidelity.

First page of Qwen-Image-Flash
Technical Report 2026Fast generation

Qwen-Image-Flash: Beyond Objective Design

T. Wu, K. Yan, Zikai Zhou, L. Jiang, J. Li, J. Zhang, K. Gao, N. Tang, S. Yin, X. Chen, et al.

Introduces a fast Qwen Image variant that emphasizes practical responsiveness while maintaining visual quality and instruction-following behavior.

First page of Adaptive Matching Distillation paper
ICML 2026Few-step generation

Optimizing Few-Step Generation with Adaptive Matching Distillation

Lichen Bai*, Zikai Zhou*, Wenliang Zhong, Shitong Shao, Shuo Yang, Shuo Chen, Bojun Cheng, Zeke Xie.

Studies how to distill generation trajectories into a small number of sampling steps through adaptive matching, targeting fast inference without sacrificing image quality.

First page of Lightning Unified Video Editing paper
ICML 2026Video editing

Lightning Unified Video Editing via In-Context Sparse Attention

Shitong Shao*, Zikai Zhou*, Haopeng Li, Yingwei Song, Wenliang Zhong, Lichen Bai, Zeke Xie.

Uses in-context sparse attention to unify video editing operations in a faster pipeline, improving edit consistency while reducing unnecessary attention computation.

First page of data-free LoRA transferability paper
ICML 2026Video diffusion

Exploring Data-Free LoRA Transferability for Video Diffusion Models

Y. Wang, W. Zhong, Lichen Bai, Zikai Zhou, Shitong Shao, Bojun Cheng, Shuo Chen, Shuo Yang, et al.

Examines whether LoRA modules can transfer across video diffusion models without access to the original training data, clarifying when lightweight adaptation remains reusable.

First page of Zigzag Diffusion Sampling paper
ICLR 2025Sampling

Zigzag Diffusion Sampling: The Path to Success Is Zigzag

Lichen Bai, Shitong Shao, Zikai Zhou, Zipeng Qi, Zhiqiang Xu, Haoyi Xiong, Zeke Xie.

Explores a non-monotonic sampling path where diffusion models can self-reflect during generation, improving sample quality through zigzag refinement.

First page of Reflective Flow Sampling Enhancement paper
TPAMI under reviewFlow sampling

Reflective Flow Sampling Enhancement

Zikai Zhou*, Muyao Wang*, Shitong Shao, Dian Xie, Lichen Bai, Zeke Xie.

Extends reflective sampling ideas to flow-based generation, targeting stronger sample quality through intermediate trajectory correction.

First page of Elucidating the Design Space of Dataset Condensation paper
NeurIPS 2024Dataset condensation

Elucidating the Design Space of Dataset Condensation

Shitong Shao, Zikai Zhou, Huanran Chen, Zhiqiang Shen.

Maps the main choices in dataset condensation and clarifies how different recipes influence downstream training performance.

First page of Similar Target Method paper
IJCNN 2024Adversarial attacks

Enhancing Adversarial Attacks: The Similar Target Method

S. Zhang, Z. Wang, Zikai Zhou, J. Liu, H. Chen.

Improves adversarial attack construction by selecting similar target classes, making perturbation objectives more effective and semantically grounded.

PDF preview unavailable
Under reviewVideo diffusion

The Blessing of Smooth Initialization for Video Diffusion Models

Shitong Shao, Lichen Bai, Zikai Zhou, Tian Ye, Yunfeng Cai, Kaishun Wu, Zeke Xie.

Studies how smooth initialization can stabilize video diffusion generation and improve temporal coherence at inference time.

* Equal contribution. Technical report entries reflect the Qwen Image series contributions listed in the CV and Google Scholar profile.

Experience

Research that ships into model systems.

Research Intern, Qwen Image Foundation Model Team, Alibaba Group

Core contributor to Qwen Image 2.0, 3.0, and Turbo. Built automated data flywheels, large-scale distillation pipelines, and pre-training evaluation systems for image foundation models.

Algorithm Engineer, Fuguang AI

Led R&D and deployment of T2V, I2V, and V2V models, including data collection, auto-labeling, quality assessment, inference optimization, and prompt-engineering modules.

Research Assistant, xLeaf Lab, HKUST(GZ)

Worked on AIGC inference acceleration and sampling optimization for diffusion and flow matching, contributing to Golden Noise and Zigzag Diffusion Sampling.

Education

HKUST(GZ) and BIT

Ph.D. student in Artificial Intelligence at HKUST(GZ), supervised by Prof. Zeke Xie. B.Eng. in Computer Science from Beijing Institute of Technology, Outstanding Graduate 2025.

Academic Service

Reviewer and chairing

ICML 2026 Gold Reviewer. Session Chair at IJCAI 2024. Reviewer for ICML, NeurIPS, ICLR, CVPR, ICCV, ECCV, AAAI, ACM MM, and IJCAI.

Talk

Inference optimization

Invited by Qingke Community to present research on inference optimization for diffusion models.