Hongyang Wei（韦弘杨）

Hongyang Wei（韦弘杨）

About Me

I’m currently a final-year M.Eng. student at Tsinghua University, supervised by Prof. Chun Yuan. Before that, I received my B.S. degree in Computer Science and Technology from Wuhan University in 2023. My research interests include unified models and visual generation.

🔥 News

  • 2026.05:  🎉 Two papers accepted at ICML 2026!
  • 2026.02:  🎉 Two papers accepted at CVPR 2026!
  • 2026.02:  🚀 Released UniRef-Image-Edit — the first to apply GRPO to multi-image composition.
  • 2026.01:  🚀 Released UniMRG — to explore how generation enhances understanding in unified models.
  • 2026.01:  🚀 Released Skywork UniPic 3.0 — the first open-source multi-image composition framework.
  • 2025.12:  🚀 Released MICo-150K — a large-scale multi-image composition dataset.
  • 2025.11:  🎉 One paper accepted at TMM!
  • 2025.08:  🚀 Released Skywork UniPic 2.0 — the first to apply GRPO to single-image editing.
  • 2025.08:  🚀 Released Skywork UniPic 1.0 — a 1.5B unified model supporting understanding, generation, and editing.
  • 2025.06:  🎉 One paper accepted at ICCV 2025!
  • 2025.03:  🚀 Released PURE — the first MLLM-based image restoration framework.

📝 Selected Publications

*indicates equal contribution

Tech Report
sym

UniRef-Image-Edit: Towards Scalable and Consistent Multi-Reference Image Editing

Hongyang Wei, et al.

Tech Report, 2026

arXiv

Tech Report
sym

Skywork UniPic 3.0: Unified Multi-Image Composition via Sequence Modeling

Hongyang Wei, et al.

Tech Report, 2026

arXiv / Code

Tech Report
sym

Skywork UniPic 2.0: Building kontext model with online rl for unified multimodal model

Hongyang Wei, et al.

Tech Report, 2025

arXiv / Code

Tech Report
sym

Skywork UniPic 1.0: Unified autoregressive modeling for visual understanding and generation

Peiyu Wang, …, Hongyang Wei, et al.

Tech Report, 2025

arXiv / Code

ICCV 2025
sym

Perceive, Understand and Restore: Real-World Image Super-Resolution with Autoregressive Multimodal Generative Models

Hongyang Wei, et al.

ICCV, 2025

Paper / Code

ICML 2026
sym

Generation Enhances Understanding in Unified Multimodal Models via Multi-Representation Generation

Zihan Su * , Hongyang Wei * , et al.

ICML, 2026

arXiv / Code

CVPR 2026
sym

MICo-150K: A Comprehensive Dataset Advancing Multi-Image Composition

Xinyu Wei * , Kangrui Cen * , Hongyang Wei * , et al.

CVPR, 2026

arXiv / Code

ICML 2026
sym

SpatialReward: Bridging the Perception Gap in Online RL for Image Editing via Explicit Spatial Reasoning

Yancheng Long, Yankai Yang, Hongyang Wei, et al.

ICML, 2026

arXiv

arxiv 2026
sym

TIIF-Bench: How Does Your T2I Model Follow Your Instructions?

Xinyu Wei, Jinrui Zhang, Zeqing Wang, Hongyang Wei, et al.

arxiv, 2026

arXiv / Code

🎖 Selected Honors and Awards

  • 2021.10: ACM-ICPC Regional Contest (Shenyang Site) Silver Medal🥈
  • 2020.10: National Scholarship at Wuhan University🏅

📖 Education

  • 2023.09 - 2026.06, M.E. in Computer Science, Tsinghua University.
  • 2019.09 - 2023.07, B.E. in Computer Science, Wuhan University.

💻 Internships