Presentation
A Few-Step Generative Model on Cumulative Flow Maps
DescriptionWe propose a unified few-step generative framework based on cumulative flow maps for long-range transport in probability space. By connecting instantaneous updates with finite-time transport, it applies broadly to diffusion and flow models, supports few-step and one-step generation, requires minimal modifications, preserves quality, and reduces inference cost across diverse tasks.
Event Type
Technical Paper
TimeWednesday, 22 July 20262:40pm - 2:50pm PDT
LocationRoom 403 B
Digital Library
PDF
Session Time & Location
Sunday, 19 July 20266:00pm - 8:45pm PDTHall K
Wednesday, 22 July 20262:00pm - 3:30pm PDTRoom 403 B
Artificial Intelligence/Machine Learning
Modeling
Full Conference Supporter
Full Conference
