New STMD method speeds diffusion model inference without teacher

Researchers have developed Stochastic Transition-Map Distillation (STMD), a novel framework designed to accelerate the inference process for diffusion models without requiring a pre-trained teacher model. This method distills the full transition map of the sampling stochastic differential equation (SDE), enabling faster, probabilistic sample generation. STMD offers a theoretical foundation with convergence bounds in Wasserstein distance and has been demonstrated on image generation tasks across MNIST, CIFAR-10, and CelebA datasets. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Accelerates diffusion model inference, potentially enabling wider use in applications requiring fast, probabilistic generation.

RANK_REASON Publication of a new academic paper detailing a novel method for accelerating diffusion model inference. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Stochastic Transition-Map Distillation (STMD)
• diffusion models
• George Rapakoulias
• MNIST
• CIFAR-10
• CelebA

• paper
• model release