PhD student, University of California, San Diego
3 papers at NeurIPS 2025
We propose HASTE, which combines holistic alignment (feature and attention) with early termination to accelerate diffusion transformer training by 28× while maintaining quality.
Generate high-performing LoRA parameters from prompts that is unseen in training.
Loong-X enables hands-free image editing using multimodal neural signals, achieving performance comparable to text-driven methods by combining BCIs with the proposed diffusion-based generative methods.
