Hello, I’m
Master’s Student at Fudan University
I am a master’s student at Fudan University working on efficient generative models for autonomous driving. My research focuses on transforming autoregressive Vision-Language-Action models into efficient parallel generative systems through discrete diffusion, flow matching, and hierarchical distillation. My recent work explores how to preserve multimodal reasoning while reducing the latency and exposure bias of sequential action generation.
Recent Updates
Selected Publications
Selected work on efficient, parallel generative models for autonomous driving. * Equal contribution. † Corresponding author.
WAM-Diff2 converts a pretrained autoregressive driving generalist into a parallel discrete diffusion VLA through progressive block-wise adaptation, block-wise distillation, and model-wise cross-scale distillation—preserving multi-task capabilities while improving inference efficiency.
Research
Parallel discrete generation and efficient decoding for Vision-Language-Action models in autonomous driving.
Progressive adaptation and hierarchical distillation that transfer capabilities from autoregressive models to diffusion systems.
Preserving scene understanding and multimodal reasoning while improving the efficiency and robustness of trajectory generation.
Experience & Education
Master’s Student
Shanghai, China
Research on efficient multimodal generative models, Vision-Language-Action models, and autonomous driving.
Information Management and Information Systems, School of Mathematics
Guangzhou, China
Undergraduate study in information management and information systems.
Contact
I am open to research discussions and collaborations on efficient multimodal generative models and autonomous driving.