👤 About Me
I am a third-year undergraduate at Xidian University, majoring in Software Engineering (Intelligent Direction). I rank 1st/335 in all Software Engineering and 1st/1342 in the Computer Category for the 2023-2024 academic year. Additionally, I am proficient in English (CET-6: 591) and French (College French Test Band 4: Excellence).
Research Interests: Multimodal LLMs Efficient AI Large Language Models Omni LLMs Token Pruning KV Cache Optimization
📖 Educations
🎓 Xidian University · Xi’an, China
B.Eng. in Software Engineering Sep. 2023 - Jun. 2027 (expected)
As a top-ranked student with a perfect GPA of 4.0/4.0, I have demonstrated consistent academic excellence in core courses such as Physics (99), Programming Design (98), Circuits (98), Discrete Math (98), Data Communications & Networks (97), and Advanced Mathematics (97).
🌴 University of California, Los Angeles · Los Angeles, USA
Visiting Student, Summer Session Jun. 2024 - Oct. 2024
Mastered mathematical modeling techniques via intensive study in Numerical Analysis.
🏫 Zhenhai High School of Ningbo · Ningbo, China
Innovation Class (Jiaochuan Academy) Sep. 2020 - Jun. 2023
Selected for the Innovation Class at Zhejiang’s top-ranked high school; 15 classmates and ~70 students school-wide are admitted to Peking University or Tsinghua University annually.
📝 Publications
Accelerating Diffusion Large Language Models with SlowFast: The Three Golden Principles
Q. Wei, Y. Zhang, Z. Liu, P. Zeng, Y. Wang, D. Liu, L. Zhang
Accepted to ICLR 2026 (CCF-A) 🎉✨
Proposed SlowFast Sampling, a dynamic strategy that adaptively alternates between exploratory and accelerated stages based on token certainty, convergence, and positional principles. Achieved 15.63x speedup on LLaDA.
📄 Paper 💻 Code
UltraHiT: A Hierarchical Transformer Architecture for Generalizable Internal Carotid Artery Robotic Ultrasonography
T. Wang*, H. Jiang*, Y. Wang*, Z. Sun, X. Yan, X. Li, G. Huang (* denotes Equal Contribution)
Accepted to ICRA 2026 (CAA-A) 🎉✨
Proposed UltraHiT, a hierarchical Transformer for autonomous robotic ultrasonography, significantly improving generalization across diverse ICA anatomies.
📄 Paper 🌐 Project Page
🚀 Projects
AudioKV: KV Cache Eviction in Efficient Large Audio Language Models
Y. Wang, P. He, X. Gui, X. Liu, J. He, X. Liu, X. Hu, L. Zhang
First Author; Submitted to ICML 2026 (CCF-A).
Proposed AudioKV, prioritizing audio-critical attention heads via semantic-acoustic alignment. Reduces memory overhead by 60% with Spectral Score Smoothing (SSS).
📄 Paper (Coming Soon)
Thinking inside the Mask: In-place Prompting in Diffusion LLMs
X. Jin, Y. Wang, Y. Gao, Z. Wen, B. Qi, D. Liu, L. Zhang
Second Author; Submitted to ACL 2026 (CCF-A).
Proposed In-Place Prompting, integrating reasoning chains directly into the mask denoising process for Diffusion LLMs.
📄 Paper 💻 Code
Self Speculative Decoding for Diffusion Large Language Models
Y. Gao*, Z. Ji*, Y. Wang, B. Qi, H. Xu, L. Zhang (* denotes Equal Contribution)
Submitted to ACL 2026 (CCF-A).
Developed SSD for Diffusion LLMs, accelerating inference via internal self-drafting and parallel verification.
📄 Paper
AudioMarathon: A Comprehensive Benchmark for Long-Context Audio Understanding and Efficiency in Audio LLMs
P. He*, Z. Wen*, Y. Wang*, Y. Wang, X. Liu, J. Huang, Z. Lei, Z. Gu, X. Jin, J. Yang, et al. (* denotes Equal Contribution)
Submitted to ACL 2026 (CCF-A).
📄 Paper 💻 Code 🗂️ Dataset
VA-Adapter: Adapting Ultrasound Foundation Model to Echocardiography Probe Guidance
T. Wang, H. Jiang, Y. Wang, Z. Sun, S. Song, G. Huang
Parameter-efficient Vision-Action Adapter for real-time probe guidance.
📄 Paper
🔥 Operating System Review (June 2025)
Open-source study guide for Xidian Software Engineering students, covering Modern Operating Systems.
📖 Read Guide
💻 Internships
🔬 EPIC Lab, School of Artificial Intelligence, Shanghai Jiao Tong University, China.
Research Assistant August 2025 - February 2026
Developed efficient algorithms for large language diffusion models, leveraging bidirectional attention as a high-performance alternative to autoregressive models. Also spearheaded KV cache eviction research for Audio LLMs.
🤖 LEAP Lab, Department of Automation, Tsinghua University, China.
Research Assistant March 2025 - September 2025
Participated in the implementation of the internal carotid artery ultrasound autonomous navigation project. My work was conducted onsite at Tsinghua University’s Central Main Building, room 601. Our work has been accepted to ICRA 2026. 🥳✨🎉
🥇 Selected Awards
🏆 2025.12: Huawei Scholarship (Top 0.1%) - Issued by Huawei Xi’an Research Institute
🏅 2023 - 2025: National Scholarship (Twice, 2023-2024, 2024-2025)
🥇 2025.04: National First Prize, National English Competition for College Students (NECCS)
🥇 2024.12: First Prize, National Undergraduate Mathematical Modeling Contest
🥇 2024.11: First Prize, 16th National Undergraduate Mathematics Competition
🥇 2024.01: National First Prize, Vocabulary Star National English Vocabulary Competition
🎖️ 2024.01: Honorable Mention, Mathematical Contest in Modeling (MCM)
🎨 Activities
🌊 Xidian Inspur Club, President (2025 - 2026)
Club Management: Directed daily operations and spearheaded club recruitment, while organizing academic workshops and orientations to grow the research community.
🧙 Hobbies
⚡ Professional Harry Potter Series Enthusiast. 8 years of deep immersion having read the books and watched the movies over 10 times; I know the Wizarding World’s evolution better than my own life trajectory (still waiting for my owl 🦉).
👯 Friends
🤝 Krysdal C. Warhol, a junior at Peking University (CS), who plans to apply for a Master’s degree in the United States. He learned Spanish and Japanese, whereas I learned French.