AI RESEARCH
PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inference
arXiv CS.AI
•
ArXi:2603.25730v1 Announce Type: cross Autoregressive video diffusion models have nstrated remarkable progress, yet they remain bottlenecked by intractable linear KV-cache growth, temporal repetition, and compounding errors during long-video generation. To address these challenges, we present PackForcing, a unified framework that efficiently manages the generation history through a novel three-partition KV-cache strategy.