Yijia Wang

Hi, I am a final-year undergraduate student at the Artificial Intelligence Lab of Southern University of Science and Technology, under the supervision of Professor Zhihai He (IEEE Fellow). I also spent an impressive summer in 2024 at Stanford University, where I deepened my understanding of algorithms and optimizations.

My research is focused on image generation, and its practical applications, including image coding and image super-resolution. I am also interested in developing efficient image generation models that offer faster speed and higher quality. My goal is to address key challenges in the image generation process, thereby improving the performance of downstream tasks that are based on generation methods.

Currently, I am seeking PhD opportunities to further explore the key issues in image generation technology. I aim to develop advanced algorithms that can significantly enhance the quality and speed of image generation, and I am eager to apply these techniques to real-world scenarios to make a positive impact.

Email  /  Google Scholar  /  Github

News

[11/04/2025] Congratulations! I have been awarded the Joyce M.Kouk Scholarship. This is a significant achievement, and I’m very grateful for the recognition of my academic efforts.

[10/27/2025] Congratulations! Our team has won the Silver Award in the 2025 China International University Students Innovation and Entrepreneurship Competition of Guangdong Province. Among the 484,500 teams that participated, only 234 teams (Top 0.07%) received this prestigious award.

[09/09/2025] Two of my patents, “Template-Replacement Image Compression & Reconstruction Method, System and Storage Medium" (CN 202511281423.X) and "Image Editing Method Based on Attention Decoupling, Device, Terminal and Storage Medium" (CN 202511281645.1), have been officially accepted by the China National Intellectual Property Administration (CNIPA).

Research Experience

(* denotes equal contributions.)

Generative Semantic Coding for Ultra-Low Bitrate Visual Communication and Analysis

Text + tiny latent steer a generative model to reconstruct images at ultra-low bitrates.

Runge-Kutta Approximation and Decoupled Attention for Rectified Flow Inversion and Semantic Editing

We improved Rectified Flow editing via a Runge-Kutta solver for accurate inversion and decoupled attention for precise control.

Understanding the Implicit User Intention via Reasoning with Large Language Model for Image Editing

We proposed CIELR for complex image editing by using LLMs to reason about implicit instructions and convert them into simple editing steps, all without fine-tuning.

Education

Scholarships/Awards

[11/2025] 2024-2025 BYD Scholarship

[09/2025] 2024-2025 Joyce M.Kouk Scholarship

[09/2025] 2024-2025 First-Class Scholarship for Outstanding Students, SUSTech

[08/2025] Silver Award in 2025 China International University Students Innovation and Entrepreneurship Competition, Guangdong

[09/2024] 2023-2024 Second-Classs Scholarship for Outstanding Students, SUSTech

[01/2024] Second Prize in the 2023 China Undergraduate Mathematical Contest in Modeling, Guangdong

[09/2023] 2022-2023 Star of Practice Scholarship, SUSTech

[09/2023] 2022-2023 First-Class Scholarship for Outstanding Students, SUSTech