Xinyu Li
Xinyu Li is currently a Ph.D. candidate in Computer Science at Fudan University, and is expected to receive his Ph.D. degree in December 2026.
He received the M.S. degree in Computer Technology from Ningxia University in 2022 and the B.S. degree in Computer Science from Qingdao Institute of Technology in 2020.
His research interests lie in Embodied AI, with a particular focus on Robot General Mobile Manipulation Policy. His previous research focused on multivariate long-term time-series forecasting and 3D face reconstruction, both of which serve as important foundations for Embodied AI. He has published seven first-author papers in conferences including ACL, WWW, ICASSP, and ICME.
He is currently seeking postdoctoral positions in Embodied AI. If you are interested, please contact him by email at 22110240025@m.fudan.edu.cn.
Publications (First-Author, Main-Paper)
Is the Attention Matrix Really the Key to Self-Attention in Multivariate Long-Term Time Series Forecasting?
ACL 2026, Oral Paper
[PDF]
Modeling Point-to-Point Dependency for High-Dimensional Long-Term Series Forecasting
WWW 2026, Oral Paper
[PDF]
MoME: Mixture of Multi-Domain Experts for Multivariate Long-Term Series Forecasting
ICASSP, 2025
[PDF]
HMSformer: Hierarchical Multi-Scale Transformer for Multivariate Long-Term Series Forecasting
ICME 2025
[PDF]
ForeNet: Unlocking Long-Term Series Forecasting in High-Dimensional Scenario via Forest Structure
ICME 2025
[PDF]
Credible and Detailed 3D Face Reconstruction in Large Pose
ICASSP 2025
[PDF]
Towards Accurate 3D Face Alignment Under Extreme Scenarios via Multi-Granularity Perturbation Relearning
ICME 2024
[PDF]
Honors and Awards
- 2025 First-Class Academic Scholarship, Fudan University
- 2024 First-Class Academic Scholarship, Fudan University
Academic Services
- Conference Reviewer: ICML, ICME
- Online Session Chair: ACL 2026