
CSD-VAR: Content-Style Decomposition in Visual Autoregressive Models
A novel content-style personalization method for Visual Autoregressive Models.
MS CS @ University of Illinois at Urbana-Champaign (UIUC)
Hi, I'm currently a Master student in Computer Science (MS CS) at University of Illinois at Urbana-Champaign (UIUC) in Fall 2025. Previously, I was an AI Research Resident at VinAI Research (acquired by Qualcomm), working on 3D generative models and 3D scene understanding mentored by Dr. Khoi Nguyen and Prof. Binh-Son Hua. I earned my B.S. in Computer Science from Ho Chi Minh University of Technology in 2023.
I'm interested in 3D vision, generative models, and computer graphics.
A novel content-style personalization method for Visual Autoregressive Models.
Introducing the Vocablulary-Free 3D point cloud instance segmentation with different solid baselines and a novel pointwise method using multimodal LLM.
A novel class-agnostic approach for 3D instance segmentation that leverages 2D mask tracking to segment 3D objects in point cloud scenes.
Tackling mode-collapses problem in existing text-to-3D generation by using personalized text prompts to diversify the joint generation.
Introducing a novel Image Prompt Score Distillation (ISD) loss to address over-smoothing and low-quality outputs in text-to-3D generation caused by mode-seeking behavior in SDS-based method.