I am Muyao Yuan (袁慕遥), a Ph.D. candidate in the MOEKLINNS Lab at Xi’an Jiaotong University, under the guidance of Prof. Weizhan Zhang. My research interests include model compression, model adaptation, and MLLM.

🔥 News

• 📅 [2025.11] 🎉 Our paper titled “InfoCLIP: Bridging Vision-Language Pretraining and Open-Vocabulary Semantic Segmentation via Information-Theoretic Alignment Transfer” has been accepted by AAAI 2026!
• 📅 [2025.05] 🎉 Our paper titled “InfoSAM: Fine-Tuning the Segment Anything Model from an Information-Theoretic Perspective” has been accepted as a Spotlight Poster at ICML 2025!

📝 Publications

InfoCLIP: Bridging Vision-Language Pretraining and Open-Vocabulary Semantic Segmentation via Information-Theoretic Alignment Transfer

M. Yuan, Y. Zhang, W. Zhang, L. Ma, Y. Gao, J. Ying, Y. Xin AAAI Conference on Artificial Intelligence (AAAI), 2026

Project Page PDF

InfoSAM: Fine-Tuning the Segment Anything Model from an Information-Theoretic Perspective

Y. Zhang*, M. Yuan*, W. Zhang, T. Gong, W. Wen, J. Ying, W. Shi International Conference on Machine Learning (ICML Spotlight), 2025

Project Page PDF Code

Adaptive Token Selection for Efficient Detection Transformer with Dual Teacher Supervision

M. Yuan, W. Zhang, C. Yan, T. Gong, Y. Zhang, J. Ying Knowledge-Based Systems (KBS), 2024

PDF Code

Lightweight Configuration Adaptation with Multi-teacher Reinforcement Learning for Live Video Analytics

Y. Zhang, W. Zhang, M. Yuan, L. Xu, C. Yan, T. Gong, H. Du IEEE Transactions on Mobile Computing (TMC), 2025

PDF

🛠 Projects

Efficient Training and Inference of Multimodal Foundation Models

Multimodal foundation models, efficient training and inference

Project Page

Lightweight and Efficient Model Design for Diverse Tasks

Model compression and efficient architecture design

Project Page

💡 Selected Patents

• Resource-Efficient Training for MLLMs via Adaptive Data Filtering
• Efficient Collaborative Inference for Autoregressive MLLMs
• Device and Edge Collaborative Personalized Inference for MLLMs with Heterogeneous Resources
• Efficient MLLM Training via Expanded Pipeline Stages
• Self-Driven Feedback and Symbolic Collaboration for MLLM Inference
• LoRA and MoE-Based Fine-Tuning for MLLMs
• Elastic Architecture Design and Pruning for MLLMs with Heterogeneous Experts
• Efficient MLLM Initialization via Weight Inheritance

🏆 Selected Awards

• Special Class Academic Scholarship, Xi’an Jiaotong University, 2025.
• Excellent Postgraduate, Xi’an Jiaotong University, 2025.
• Outstanding Graduate, Xi’an Jiaotong University, 2022. [Certificate]
• First Prize Scholarship, Xi’an Jiaotong University, 2020. [Certificate]
• National Second Prize, WeChat Mini Program Development Competition, 2020. [Certificate]
• Samsung Scholarship, 2019. [Certificate]
• National Second Prize, Contemporary Undergraduate Mathematical Contest in Modeling (CUMCM), 2019. [Certificate]
• First Prize, Chinese Physics Olympiad (CPhO), 2017. [Certificate]

📖 Educations

• 2022.09 – now, Xi’an Jiaotong University — PhD Candidate in Computer Science
  • Supervisor: Prof. Weizhan Zhang

• 2018.09 – 2022.06, Xi’an Jiaotong University — B.S. in Automation
  • GPA: 4.0/4.3 (Ranked 1st out of 180) [Transcript]

🏢 Internships

• 2023.10 – 2024.11, China Telecom Artificial Intelligence Technology, Beijing.
• 2025.10 – 2025.11, E-surfing Vision Technology, Shanghai.

🎓 Academic Service

• Reviewer for CVPR, AAAI, TMC, Neural Computation, Neural Networks.