Wenjie Du

Welcome to my homepage!🤗 I'm Wenjie Du (杜文杰), a research intern in ENCODE Lab of Prof. Huan Wang at Westlake University. I obtained B. Eng. in Computer Science and Technology from Sichuan University in 2024 and have received the National Scholarship in 2023. I was fortunate to work as a research assistant at HKUST in 2025, a research intern at the Institute for AI Industry Research (AIR), Tsinghua University in 2024, and a software developer intern at ByteDance in 2023. I am grateful to all my mentors, supervisors, and friends for their kindness and support throughout my journey!

Email  /  CV  /  Scholar  /  Linkedin  /  Github

I am seeking PhD positions starting Fall 2026.

profile photo

Research

My research interests lie in large language models and machine learning systems. I focus on efficient reasoning in LLMs through understanding their inherent working mechanisms. I am also working towards efficient, robust, and reliable multi-agent systems for complex real-world tasks.

clean-usnob Which Heads Matter for Reasoning? RL-Guided KV Cache Compression
Wenjie Du, Li Jiang, Keda Tao, Xue Liu, Huan Wang,
arXiv, 2025
page / arXiv

An RL-based framework that identifies reasoning-critical attention heads in LLMs to enable KV cache compression, achieving 20-50% reduction in cache size with minimal performance loss.

clean-usnob AutoDroid-V2: Boosting SLM-based GUI Agents via Code Generation
Hao Wen, Shizuo Tian, Borislav Pavlov, Wenjie Du, Yixuan Li, Ge Chang, Shanhui Zhao, Jiacheng Liu, Yunxin Liu, Ya-Qin Zhang, Yuanchun Li,
MobiSys, 2025   (Best Artifact Award)
code / arXiv / pdf

A document-centered framework that converts mobile UI automation into code generation, enabling the script-based mobile GUI agent to achieve higher success rates and lower latency than step-wise agents.

clean-usnob LLM-Explorer: Towards Efficient and Affordable LLM-based Exploration for Mobile Apps
Shanhui Zhao, Hao Wen, Wenjie Du, Cheng Liang, Yunxin Liu, Xiaozhou Ye, Ye Ouyang, Yuanchun Li,
MobiCom, 2025
code / arXiv

An efficient mobile app exploration agent that uses LLMs for knowledge maintenance rather than action generation, achieving the highest coverage with 148x lower cost than baselines.


This webpage is built upon the source code of Jon Barron.