Hi there, I'm Kaiyu Wu (伍开钰), and I go by Marco. 👋

🌍 Real-world Agents | 🤖 Autonomy | 🧬 Self-Evolution

Homepage GitHub Google Scholar Email

🧐 About Me

I am a Master’s student (2025–2028) at the School of Computer Science and Engineering, Sun Yat-sen University (SYSU), and a member of the HCP Lab, advised by Prof. Keze Wang.

🌱 Status: I am actively looking for Research Collaborations and Internship Opportunities. Feel free to reach out!

🎓 Research Interests

My primary research centers on multimodal post-training and agents, with a focus on building reliable supervision and evaluation signals: trajectory supervision for latent visual search in GUI grounding, consistency rewards for multimodal reasoning, and rubric-based evaluation for open-ended agent reports.

🎯 Ultimate Goal: Agents that exist in the real world as independent individuals, completing tasks autonomously and continually evolving themselves.

📝 Selected Publications

  • Recurrent Latent Visual Search for GUI Grounding
    Kaiyu Wu, Beichen Zheng, Weiyao Huang, Keze Wang
    [Paper Link]

  • [ICASSP 2026] Weather-R1: Logically Consistent Reinforcement Fine-Tuning for Multimodal Reasoning in Meteorology
    Kaiyu Wu, Pucheng Han, Hualong Zhang, Naigeng Wu, Keze Wang
    [Paper Link] [Code]

  • [EMNLP 2026 Findings] End-to-End Differentiable Optimization of Tool-Augmented Visual Reasoning via Probabilistic Graphs
    Wentao Wan, Kaiyu Wu, Qingyang Ma, Nan Kang, Yunjie Chen, Liang Lin, Keze Wang
    [Paper Link] [Code]

For a full list of publications, please visit my Google Scholar.

💬 Connect with Me