AI RESEARCH

Accelerating Rectified Flow Models via Trajectory-Aware Caching

arXiv CS.CV

ArXi:2605.16789v1 Announce Type: new Diffusion and rectified flow (RF) models generate high-fidelity images and videos, but their iterative velocity-field evaluations are computationally expensive. Existing caching methods accelerate sampling by skipping timesteps, yet their coarse approximations