Applying for PhD programs · Fall 2027. Always happy to have a coffee chat or explore collaborations — feel free to reach out!

Yuyang Peng 彭雨洋

I build agents that see the live world.

CS undergraduate at Huazhong University of Science and Technology, currently a research intern with Prof. Jieyu Zhao at USC.

See my work Get in touch

agent@yuyang — ~/homepage

📄Publications3 papers + 1 project 🧪Research3 labs 🏅HonorsICPC · NOI ✉️Contactsay hi

About

📍 Wuhan, China

Hi! I'm Yuyang, a fourth-year Computer Science (Turing Class) undergraduate at HUST. I'm curious about how machines can perceive a constantly changing world and act in it — from answering questions about today's news images, to coordinating fleets of LLM agents, to operating real desktop software.

Right now I'm working on computer-use agents: synthesizing GUI training data at scale and using reinforcement learning to teach agents to complete complex, multi-step tasks. Before that I competed in algorithmic programming (NOI, ICPC, CCPC), which still shapes how I think about systems and efficiency.

Multimodal Learning Computer Vision LLM Agents Computer-Use Agents Reinforcement Learning

0papers · plus 1 open-source project

0first / co-first author

0research internships

0contest medals & awards

Research Journey

Feb 2026 — Present● Now

Computer-Use Agents

Research Intern · Prof. Jieyu Zhao · University of Southern California
- GUI synthetic data generation and reinforcement-learning training techniques for computer-use agents.
- Building automated synthesis pipelines that improve agent policy learning on complex GUI tasks.
Aug 2025 — Feb 2026KDD 2026 · First author

CoAct: Parallel LLM Agents

Research Intern · Prof. Qinbin Li · HUST
- Proposed an online scheduling framework based on contrastive task allocation.
- Minimizes inter-node communication overhead and improves parallel execution efficiency of LLM agents.
Dec 2024 — Aug 2025NeurIPS 2025 · Co-first author

LiveVQA: Live Visual Knowledge

Research Intern · Prof. Yao Wan · HUST
- Constructed a large-scale open dataset for evaluating up-to-date visual understanding.
- Developed and released an evaluation toolkit to ensure reproducibility.
Sep 2023 — Jun 2027 (expected)Education

B.E. in Computer Science (Turing Class)

Huazhong University of Science and Technology · Wuhan