Hi, Nice to meet you! My name is Yu He. I’m currently a third-year undergraduate student at School of Intelligent Software and Engineering, Nanjing University (NJU). I’m fortunate to be advised by Prof. Ke Xu, and to collaborate closely with Dr. Wenchao Li. My time working with them has been both memorable and formative for my research journey.
My research interests lie at the intersection of Data-centric AI, Interactive Visualization, and Human-AI Interaction. I study how interaction can turn data into a steering wheel for reliable AI–improving models through explainable, iterative data decisions.
More specifically, I am interested in:
- Interactive data-centric workflows for model improvement: designing explainable, human-in-the-loop interfaces that support data diagnosis to iteratively enhance model performance.
- Visual sensemaking for complex structures: building visually guided systems that help users explore structrued relational data (e.g., networks) as well as unstructured / multimodal data (e.g., text, images, videos) to surface potential patterns and actionable insights.
- Reliability over distribution shift and time: studying how data decisions and time-aware learning affect robust generalization, and developing methods that make these trade-offs measurable and controllable.
I’m actively seeking potential research internship opportunity in the United States, Mainland China and Hong Kong. I’m also open to Research Collaborations! If my background aligns with your interests or if you have positions for Summer Research Intern 2026 or any interesting topics, I would be very happy to connect. Feel free to reach out to me, let’s go together to Make the Impossible Possible and the Possible Easy.
🔥 News
- 2026.01: 🎉🎉 Our collaborative work NetworkCanvas has been conditionally accepted by CHI 2026. We also submitted two papers on LLM continual learning and visualization for multimodal understanding to ICML 2026 and DIS 2026, respectively. Many thanks to all collaborators!
- 2025.09: 🎉🎉 One collaborative papers on Data Exploration and Data Narrative Generation have been submitted to CHI 2026. Many thanks to my wonderful collaborators!
- 2025.03: 🎉🎉 I start my journey in NJU iMATE Lab as a research intern!
📝 Publications
NetworkCanvas: Supporting Progressive Network Visualization Exploration via Adaptive Recommendations
Wenchao Li, Yuewen Gao, Yu He, Cong Zhu, Ke Xu
NetworkCanvas is an adaptive, AI-assisted exploration system for complex networks. It combines interactive visualization, recommendation guidance, and a traceable exploration history tree to support progressive, explainable network analysis.
Multimodal Analysis of News Videos: Event-based Visual Summarization over Serial Broadcasts
Yuewen Gao, Yu He, Xianglei Lv, Yingying Dong, Ke Xu
NewsTract is a multimodal news intelligence system that segments video news into atomic events, extracts key semantic signals, and models event relations. It provides an interactive visualization interface for exploring complex news narratives and uncovering hidden event connections.
MSSR: Memory-Aware Adaptive Replay for Continual LLM Fine-Tuning
Yiyang Lu, Yu He, Jianlong Chen, Hongyuan Zha
MSSR is a cognitive-inspired continual learning framework that models LLM memory strength based on a survival-form forgetting curve. By scheduling adaptive replay for vulnerable samples, MSSR effectively mitigates catastrophic forgetting in online fine-tuning scenarios.
🎖 Selected Honors and Awards
- 2025.11 Nanjing University People’s Scholarship.
- 2025.05 Meritorious Winner (International Second Prize), MCM/ICM (Top ~7% teams worldwide).
📖 Educations
- 2023.09 - now, Software and Engineering, School of Intelligent Software and Engineering, Nanjing University.
💻 Internships
- 2025.03 - now, Research Intern, iMATE Lab advised by Prof. Ke Xu, Nanjing University.