Xinting Liao Headshot

Xinting Liao

Vector Distinguished Postdoctoral Fellow

Website Google Scholar

Xinting Liao is a Postdoctoral Fellow at the Vector Institute, where she is advised by Dr. Xiaoxiao Li and Dr. Deval Pandya. Her current research focuses on federated learning on foundation models and out-of-distribution (OOD) robustness. She serves as the reviewer on NeurIPS, ICML, ICLR, AAAI, CVPR, ACM MM, TPAMI, and ACL ARR.

She earned her Ph.D. from Zhejiang University in June 2025, where she was supervised by Prof. Xiaolin Zheng and Prof. Chaochao Chen. During her doctoral studies, she participated in a joint Ph.D. program at the National University of Singapore under the supervision of Prof. Tat-Seng Chua and Wenjie Wang. Her earlier work explored trustworthy machine learning, with a specific emphasis on federated learning with non-IID data, privacy-preserving algorithms, and cross-domain recommendation systems.

Research Interests

  • Federated Learning on Foundation Models
  • Out-of-distribution Robustness
  • Trustworthy Machine Learning

Highlights

  • Addressed Data Heterogeneity: Developed methods (HyperFed and FedRANE) to mitigate non-IID data heterogeneity using hyperbolic spaces and graph attention mechanisms (IJCAI ’23 and ACM MM ’23).
  • Solved Representation Collapse: Proposed FedU2 (CVPR ’24) to prevent representation collapse in unsupervised federated learning.
  • Pioneered Pareto-Optimal Aggregation: Explored a Pareto-optimal aggregation framework to balance local personalization and global generalization objectives (IJCAI ’23, ACM MM ’23, and CVPR ’24).
  • Handled Out-of-Distribution Shifts: Conducted research on FOOGD (NeurIPS ’24) and FOCoOp (ICML’25) to generalize federated models to handle OOD data shifts in real-world deployments.

Publications

More Publications