About me

I am a second-year Computer Science Ph.D. student at UIUC, where I am advised by Prof. Jiawei Han and co-advised by Prof. Minjia Zhang. Before UIUC, I received my master’s degree from CMU, where I was fortunate to work with Prof. Carlee Joe-Wong, and I received my undergraduate degree from Davidson College.

My research asks: what does it take to build long-horizon agent systems we can trust with real work? I approach this from three angles:

• Agent memory. I build memory systems and benchmarks that help agents track evolving state, remembering what matters and updating what’s changed across sessions.

• Efficient agent systems. I take a systems view of the agent stack, studying what to keep, compress, and recompute so that reliability scales without cost growing with it.

• Agent evaluation and trustworthy behavior. I work on making agent behavior measurable and failures traceable in controlled environments, from a single agent’s actions to entire agent networks, so that trust is built in, not bolted on.

Feel free to reach out at xfan31 [at] illinois.edu if you are interested in any discussion or collaboration!

News

• 2026-08 New preprint: Can Agent Memory Systems Track Evolving State?
• 2026-08 Presented our Blue Sky Ideas Track paper Trustworthy Agent Network as an oral at KDD 2026 in Jeju!
• 2025-12 Presented both our FedLink and FedGraph papers at NeurIPS 2025 @NPGML.
• 2025-08 Started my Ph.D. at UIUC.

Publications

* denotes equal contribution

Preprints

• Can Agent Memory Systems Track Evolving State?
  Xinyi Fan*, Miri Liu*, Ruozhen Yang, Siru Ouyang, Jiawei Han
  arXiv preprint, Aug 2026.
  [paper]

2026

• Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted On
  Yixiang Yao, Yuhang Yao, Xinyi Fan, Jiechao Gao, Jie Wang, Minjia Zhang, Srivatsan Ravi, Carlee Joe-Wong
  In Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), Blue Sky Ideas Track, Aug 2026. Oral.
  [paper]

2025

• Federated Link Prediction on Dynamic Graphs
  Yuhang Yao*, Xinyi Fan*, Ryan A. Rossi, Sungchul Kim, Handong Zhao, Tong Yu, Carlee Joe-Wong
  In New Perspectives in Graph Machine Learning (NPGML) Workshop at NeurIPS, Dec 2025.
  [paper]

• FedGraph: A Research Library and Benchmark for Federated Graph Learning
  Yuhang Yao, Yuan Li, Xinyi Fan, Junhao Li, Kay Liu, Yu Yang, Weizhao Jin, Srivatsan Ravi, Philip S. Yu, Carlee Joe-Wong
  In New Perspectives in Graph Machine Learning (NPGML) Workshop at NeurIPS, Dec 2025.
  [paper]