I am currently a Research Assistant at the Multimedia Laboratory (MMLab), The Chinese University of Hong Kong (CUHK), working with Prof. Xiangyu Yue. Before that, I received my Bachelor’s degree in Software Engineering from the Harbin Institute of Technology, where I was advised by Prof. Zhiying Tu.

My research interests include Multimodal Large Language Models (MLLMs) and Agentic AI. For MLLMs, I focus on multimodal architecture design and modality fusion across heterogeneous signals. For Agentic AI, I am interested in LLM-driven agents for real-world tasks, with a focus on agentic workflow design and post-training.

📝 Publications

Under Review

Beyond Omni: A Unified MLLM for Embodied Sensory Understanding

Shiyun Xiong, Dongming Wu, Peiwen Sun, Bokang Yang, Jiaxu Wang, Xiaolou Sun, Mingxing Wen, Ruiqing Yin, Wang Geng, Xiao-Hui Li, Xiangyu Yue

arXiv preprint arXiv:2610.26751

• A unified MLLM that goes beyond seeing and hearing, integrating diverse sensory inputs for physical-world understanding.

Under Review

Benchmark Everything Everywhere All at Once

Shiyun Xiong, Dongming Wu, Peiwen Sun, Yuang Ai, Bokang Yang, Wencheng Han, Xiao-Hui Li, Xiangyu Yue

arXiv preprint arXiv:2606.06462

• An autonomous agent that builds evaluation benchmarks from natural-language goals, covering task design, data annotation, and quality control.

IEEE TSC

User-Oriented Fuzzy Requirement Clarification for Services Delivery in the Metaverse

Bolin Zhang, Dianhui Chu (Member, IEEE), Shiyun Xiong, Qianjun Gong, and Zhiying Tu (Member, IEEE)

IEEE Transactions on Services Computing, 2025

• A user-oriented framework that clarifies ambiguous service requirements through multi-turn dialogue, personalized questioning, and user simulation.

arXiv

Guiding VLM Agents with Process Rewards at Inference Time for GUI Navigation

Zhiyuan Hu, Shiyun Xiong, Yifan Zhang, See-Kiong Ng, Anh Tuan Luu, Bo An, YAN Shuicheng, Bryan Hooi

arXiv preprint arXiv:2504.16073

• Inference-time process reward guidance helps VLM agents select better actions for GUI navigation without fine-tuning the agents.

EMNLP

BehaviorSFT: Behavioral Token Conditioning for Health Agents Across the Proactivity Spectrum

Yubin Kim, Zhiyuan Hu, Hyewon Jeong, Eugene Park, Shuyue Stella Li, Chanwoo Park, Shyun Xiong, MingYu Lu, Hyeonhoon Lee, Xin Liu, Daniel McDuff, Cynthia Breazeal, Samir Tulebaev, Hae Won Park

Findings of the Association for Computational Linguistics: EMNLP 2025

• Behavioral token conditioning enables health agents to adapt between reactive responses and proactive assistance across clinical scenarios.

ICWS

RealWeb: A Benchmark for Universal Instruction Following in Realistic Web Services Navigation

Bolin Zhang, Shiyun Xiong, Dianbo Sui, YunZhe Xu, Zhiying Tu, Dianhui Chu

2024 IEEE International Conference on Web Services

• A benchmark for evaluating how web agents follow diverse natural-language instructions to navigate realistic web services.

🧑‍💻 Experience

• 2025.09 - Now, Research Assistant, Multimedia Laboratory, The Chinese University of Hong Kong.
• 2024.07 - 2025.02, Research Intern, Institute of Data Science/School of Computing, National University of Singapore.
• 2023.10 - 2024.07, Research Intern, Research Center of Intelligent Computing for Enterprises & Services, Harbin Institute of Technology.

💼 Services

• 2025, Reviewer: ICLR 2025 Workshop LLM Reason and Plan, ACL ARR 2025 February

🎖️ Honors and Awards

• 2021-2024, The People’s Scholarship in China
• 2023.09, Provincial Pirst Prize, China Undergraduate Mathematical Contest in Model (CUMCM)

⚙️ Skills:

• C/C++, Java, Python, Pytorch, SQL, Selenium, Android Studio, etc.