6 papers
A Quantitative Approximation Framework for Flow Distillation in Diffusion Models
Weiguo Gao, Ming Li, Lei Shi +1
We develop a quantitative framework for diffusion distillation by viewing few step sampling as approximation through compositions of learned flow maps. For trajectory distillation…
Toward Theoretical Insights into Diffusion Trajectory Distillation via Operator Merging
Weiguo Gao, Ming Li
Diffusion trajectory distillation accelerates sampling by training a student model to approximate the multi-step denoising trajectories of a pretrained teacher model using far fewe…
ProFlow: Zero-Shot Physics-Consistent Sampling via Proximal Flow Guidance
Zichao Yu, Ming Li, Wenyi Zhang +2
Inferring physical fields from sparse observations while strictly satisfying partial differential equations (PDEs) is a fundamental challenge in computational physics. Recently, de…
Terminally constrained flow-based generative models from an optimal control perspective
Weiguo Gao, Ming Li, Qianxiao Li
We address the problem of sampling from terminally constrained distributions with pre-trained flow-based generative models through an optimal control formulation. Theoretically, we…
Tree Reward-Aligned Search for TReASURe in Masked Diffusion Language Models
Zichao Yu, Ming Li, Wenyi Zhang +1
Tree search has recently emerged as a powerful framework for aligning generative models with task-specific rewards at test time. Applying tree search to Masked Diffusion Language M…
Convergence Dynamics and Stabilization Strategies of Co-Evolving Generative Models
Weiguo Gao, Ming Li
The increasing prevalence of synthetic data in training loops has raised concerns about model collapse, where generative models degrade when trained on their own outputs. While pri…