Hi there 
Welcome to my Homepage!
I am a M.S. student at Shenzhen University, focusing on Embodied AI, World Models, VLA/VLN, and Multimodal Decision-Making. I have research experience in action world models, reinforcement learning, and large model training & fine-tuning.
Feel free to reach out if you are interested in collaboration or potential opportunities.
News
- 2025.09 🚀🚀 Started working on SLAW-Nav: world model-based counterfactual reasoning for VLN.
- 2025.07 🎉🎉 Started my research internship at Huawei on World Models for Embodied AI.
- 2025.03 🚀🚀 Started working on MAPF-World: Action World Model for Multi-Agent Path Finding.
- 2024.07 🎉🎉 Started my M.S. studies at Shenzhen University.
Experience
Huawei Technologies Co., Ltd.
2025.07 - 2025.10
World Model Research Intern · Shenzhen, China
Research on predictive decision-making mechanisms of world models for embodied AI and VLA scenarios.
2025.07 - 2025.10
World Model Research Intern · Shenzhen, China
Research on predictive decision-making mechanisms of world models for embodied AI and VLA scenarios.
Shenzhen University
2024.07 - Present
M.S. in Computer Science · GPA: 3.5/4
Research interests: Embodied AI, World Models, VLA/VLN, Multimodal Decision-Making.
2024.07 - Present
M.S. in Computer Science · GPA: 3.5/4
Research interests: Embodied AI, World Models, VLA/VLN, Multimodal Decision-Making.
Shenzhen Technology University
2019.09 - 2023.06
B.E. · GPA: 4.0/4.5
2019.09 - 2023.06
B.E. · GPA: 4.0/4.5
Publications
(* equal contribution · † corresponding author)
cVAN: A Novel Sleep Staging Method via Cross-View Alignment Network
Z. Yang et al.
Published in IEEE Journal of Biomedical and Health Informatics (JBHI), vol. 29, no. 7, pp. 4659-4671, July 2025.
IEEE JBHI 2025 [paper]
Z. Yang et al.
Published in IEEE Journal of Biomedical and Health Informatics (JBHI), vol. 29, no. 7, pp. 4659-4671, July 2025.
IEEE JBHI 2025 [paper]
- IEEE JBHI 2025 cVAN: A Novel Sleep Staging Method via Cross-View Alignment Network
[paper]
Projects
SLAW-Nav: Sparse Latent Action World Model for Vision-Language Navigation
2025.09 - Present
Proposed a world model-based counterfactual reasoning framework for VLN, enabling agents to imagine future outcomes around candidate actions. Designed skip-step sparse latent prediction with Qwen2.5-VL, achieving 62.0% SR / 56.0% SPL on R2R-CE.
2025.09 - Present
Proposed a world model-based counterfactual reasoning framework for VLN, enabling agents to imagine future outcomes around candidate actions. Designed skip-step sparse latent prediction with Qwen2.5-VL, achieving 62.0% SR / 56.0% SPL on R2R-CE.
MAPF-World: Action World Model for Multi-Agent Path Finding
2025.03 - 2025.08
Co-designed an action world model unifying action generation and environment dynamics modeling via a Transformer-based fast-slow dual-system architecture with Spatial Relational Encoding (SRE).
2025.03 - 2025.08
Co-designed an action world model unifying action generation and environment dynamics modeling via a Transformer-based fast-slow dual-system architecture with Spatial Relational Encoding (SRE).
Awards
- 2023, First Prize, “Challenge Cup” National Competition — Fatigue Driving Recognition (Huawei Cloud Track).
- 2022, National First Prize, China Undergraduate Mathematical Contest in Modeling (CUMCM).
Services
- Reviewer: ICML.