About Me

I am a Joint PhD student at USTC & Shanghai AI Lab. I received my bachelor's degree from Tongji University in 2023. My earlier research focused on low-level computer vision, including image restoration and video super-resolution, under the guidance of Prof. Chao Dong.

During my internship at Shanghai AI Lab, I worked with Dr. Yihao Liu and Prof. Yu Qiao on multimodal image understanding and unified models for image generation and understanding, with a focus on perceptual-level image understanding and assessment.

I am currently a Qwen Image Research Intern at Alibaba Tongyi Lab, focusing on post-training for image generation models.

Experience

Alibaba Tongyi Lab

Qwen Image Research Intern

Aug. 2026–Present

Shanghai AI Lab

Research Intern

Jul. 2024–Jul. 2026

News

• Jul. 30, 2026 I started a research internship with the Qwen Image team at Alibaba Tongyi Lab, focusing on post-training for image generation models.
• May, 2026 We release StableI2I, a fidelity-oriented evaluation framework for image-to-image transition that diagnoses unintended changes across semantic consistency, structural fidelity, and low-level appearance. [Paper] [Project] [Code] [Checkpoint] [Benchmark]
• May, 2026 Three papers accepted by ICML'26. UniPercept was selected as a Spotlight paper.
• Feb, 2026 Two papers accepted by CVPR'26.
• Jan, 2026 Two papers accepted by ICLR'26.
• Dec, 2025 We released UniPercept, a unified framework addressing the limitations of current multimodal LLMs in perceptual-level image understanding, specifically across aesthetics, quality, structure, and texture. The release features UniPercept-Bench, a comprehensive benchmark supporting both Visual Rating (VR) and Visual Question Answering (VQA) for IAA/IQA/ISTA tasks, alongside a generalizable baseline model. Beyond evaluation, UniPercept functions as a robust reward model for post-training text-to-image systems and serves as a perceptual diagnostic tool for analyzing datasets and model outputs. [Paper] [Project] [Code] [Checkpoint] [Benchmark]
• Sep, 2025 We released ArtiMuse, a MLLM for professional aesthetic understanding, which is trained on ArtiMuse-10K, a meticulously curated, expert-annotated dataset. ArtiMuse-10K systematically defines eight explainable and fine-grained aesthetic attributes (e.g., Composition & Design, Visual Elements & Structure, ...), with a wide coverage of diverse visual domains (e.g. Graphic Design, AIGC-generated Images, ...). ArtiMuse was officially released at WAIC 2025, in the forum "Evolving with AI: The Iteration and Resilience of Artistic Creativity". [Online Demo v1.0] [Paper] [Project] [Code] [Checkpoint] [Dataset]