Abstract: Urban traffic flow generation is a critical component of urban planning and management. However, existing models are often insufficient in capturing complex spat-temporal and semantic ...
This repository provides the official implementation of the paper VITA: Vision-to-Action Flow Matching Policy (ICLR 2026). VITA is a noise-free, conditioning-free policy learning framework that learns ...
Abstract: Although diffusion models in text-to-speech have become a popular choice due to their strong generative ability, the intrinsic complexity of sampling from diffusion models harms their ...