Personal Website
Qiwen Xiao
a.k.a. Charley.
CompSci undergraduate at SUSTech since 2022.
Passionate about frontiers of AI.
Selected Research
See all
2024
Undergraduate Researcher, AI Research Institute @ Shenzhen MSU-BIT University
Developed as first author StyleMeta, a meta network combining flow matching with dynamic parameter generation for efficient multi-style graphical layout generation.
2023
Member, Supercomputing Team @ SUSTech
Participated in major supercomputing competitions, exhibited at SC24, and co-organized the SUSTech SuperComputing Competition.
2025
Undergraduate Research Assistant, Computer Vision Lab @ UC Irvine
Used XGBoost, Levenberg–Marquardt, and the PROSPECT leaf model to study the relationship between radiance spectrum and vegetation water thickness.
Selected Projects
See all
2026
Physics-Guided Conditional Flow Matching for Lensless Imaging
A conditional generative approach to diffuser-based lensless reconstruction that improves over ADMM and supervised U-Net baselines, with x-prediction performing best in our study.
2025
ODE Samplers for Rectified Flow Generative Models
A systematic comparison of six ODE samplers for rectified flow on EuroSAT, showing that Euler is suboptimal and that multi-step or adaptive solvers provide a better quality–efficiency trade-off.
2025
Boston in 3D: A Citywide Object Archive Using Nerfstudio
Led a team to build a full pipeline for capturing objects across Boston, reconstructing them with Nerfstudio, and presenting them through an interactive map-based web interface.
2024
Explainable Writing Style Detection
Built an interactive authorship attribution system that combines fastText with SHAP and LIME, achieving strong multi-author classification performance while making predictions interpretable to end users.