Education
Sun Yat-sen University,
Guangzhou·China
09/2021 — Present
B.S. in Computer Science, GPA 4.23 / 5.00 Rank 3 / 291 (Top 1%)
Publications
* denotes equal contribution
-
(ICLR 2025 Spotlight) AgentTrek: Agent Trajectory
Synthesis via Guiding Replay with Web Tutorials.
Yiheng Xu*, Dunjie Lu*, Zhennan Shen*, Junli Wang, Zekun Wang, Yuchen Mao, Caiming Xiong, Tao Yu
Page · PDF -
(ICML 2025 Poster) Aguvis: Unified Pure Vision
Agents for Autonomous GUI Interaction.
Yiheng Xu*, Zekun Wang*, Junli Wang*, Dunjie Lu, Tianbao Xie, Amrita Saha, Doyen Sahoo, Tao Yu, Caiming Xiong
Page · PDF -
(Neurips 2025 UnderReview) OpenCUA: Open
Foundations for Computer-Use Agents.
Xinyuan Wang*, Bowen Wang*, Dunjie Lu*, Junlin Yang*, Tianbao Xie*, Junli Wang* ... Tao Yu
Page · PDF
Research Experience
Research Assistant to Prof. Tao Yu
07/2024 – Present
Topic: Multimodal Computer-Use Agents
Projects
OSWorld:
Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer
Environments
06/2024 – Present
- Established a comprehensive benchmark framework with 369 diverse real-world tasks in an authentic operating system environment.
- Engineered a scalable Docker-based testing infrastructure with parallel execution capabilities, achieving 8x throughput improvement in agent evaluation.
Internships
Alibaba Qwen Team, Beijing·China
05/2025 – Present
- Constructing Computer-use Agent System.
Selected Awards and Honors
Scholarships
- National Scholarship — 12/2022
- SYSU Outstanding First-Class Scholarship — 12/2022, 12/2024
- SYSU Outstanding Second-Class Scholarship — 12/2023
Awards
- First Prize, China Undergraduate Mathematical Contest in Modeling — 12/2023
Additional Information
Research Interests
- Natural Language Processing – Large Language and Vision-Language Models
- Computer-Use Agents – Vision-Language models for real-world computer tasks
Skills
- Deep Learning: PyTorch, Megatron-LM, vLLM, DeepSpeed
- Programming Languages: Python, C++