Yuzhang Zhong

Yuzhang Zhong

Visiting predoctoral fellow at Northwestern University.
Interested in Human-Agent Interaction!

Yuzhang Zhong 钟宇璋 /juːˈʒɑːŋ dʒʊŋ/

🎆 Welcome to my homepage.

I am a visiting predoctoral fellow at Northwestern University, where I am advised by Prof. Stephen Xia. I study Computer Science at the University of Macau (Honours College), expect to graduate in 2027, and spent Spring 2026 as an exchange student at the University of Illinois Urbana-Champaign. I also collaborate with Prof. Sherry Wu's WInE Lab at Carnegie Mellon University.

Research Interests

My research focuses on Human-Agent Interaction. I study how agents choose their next steps during a task and whether differences between runs show where people should step in. I also study how an agent represents a person’s decisions and how that person can correct the model.

Understanding Agent Behavior

  • TraceFork [EMNLP 2026 Demo] aligns GUI-agent trajectories from different runs of the same task. It uses each action and its page context to match steps even when the agents navigate differently. Researchers can examine where the paths diverge and whether they later rejoin. They can then judge whether a divergence looks like a mistake or another way to reach the goal.
  • RewindAI [UIST 2026 Poster] uses an autobiographical AI clone to model how a person makes social decisions. Users can inspect the experiences and relationships behind the clone’s responses, then revise parts that feel wrong. They can also simulate a path they did not take to reflect on that choice.

Learning with AI Agents

  • Dream Factory turns textbook stories into role-play with LLM agents. Learners play one character while agents play the others. The agents respond in character, so learners can explore why those characters act as they do within the story’s historical setting.

Other Projects

Embodied Human-Agent Interaction

These projects look at how human intent shapes an agent's actions in physical settings.

  • SNS-Grasp uses semantic information about an object to guide robotic grasp generation. It favors grasps that fit what the person plans to do with the object.
  • Flexible Smart Glasses lets wearers decide which sensors the AI uses. The project explores how that choice affects robot control and awareness of bodily signals.

News

Beyond Research

✈️ Outside research, I enjoy travelling, board games, singing, and movies. I have visited the United States, Canada, Singapore, Malaysia, and Thailand.

🌶️ I am from Chengdu, and I love spicy food. Fellow spicy-food lovers are always welcome to reach out.

📮 I am also a RedNote creator, sharing pieces of life, romance, reflection, and projects I build, with 42k+ likes and favorites.

I Am Looking For

I am always looking for ambitious professors, advisors, and collaborators who enjoy building thoughtful AI systems with a human-centered spirit. I am also happy to talk with people from the humanities, psychology, economics, and other fields who are interested in Human-Agent Interaction. Please feel free to reach out if our interests overlap.

Other Ideas

I also enjoy exploring small humanities-oriented ideas and creative tools.

  • Dream of the Red Chamber Emotional Landscape - A digital humanities project for reading Dream of the Red Chamber through space, characters, and emotional heat across 120 chapters, selected as an excellent work for Zhihu’s Seeing Humanity Anew with AI event.
  • Stray Cats Simulator - A RedNote stray-cat life simulator with 700k+ users; ranked #3 among the platform’s mini-games.
  • Movie Companion.skill - A spoiler-free Codex movie companion that stays with the current moment instead of revealing what comes next, and one of the earliest REDskills on RedNote.

These ideas are not directly tied to my main research agenda, but I enjoy exploring them. Please feel free to reach out if any resonate with you.

Views --