arXiv Machine Learning

Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms

arXiv AI
Aug 3

Demystifying Video Reasoning

arXiv:2603. 16870v3 Announce Type: replace-cross Abstract: Recent advances in video generation have revealed an unexpected phenomenon: diffusion-based video models exhibit non-trivial reasoning capabilities.

By Ruisi Wang, Zhongang Cai, Fanyi Pu, Junxiang Xu, Wanqi Yin, Maijunxian Wang, Ran Ji, Chenyang Gu, Bo Li, Ziqi Huang, Hokin Deng, Dahua Lin, Ziwei Liu, Lei Yang
arXiv Computer Vision
Oct 22

Rethinking Video Autoencoders for Interactive World Models

The paper introduces CADRE, a family of causal diffusion representation autoencoders designed for video world models that balance spatio‑temporal compression. By training the same world model across different CADRE variants in manipulation and navigation tasks, the authors map the latency‑quality trade‑off and uncover that strong compression favors diffusion decoders, temporal compression is more effective in the autoencoder for manipulation, and spatial compression is preferable when the camera moves. The study also reveals that coarse spatial tokens degrade motion perception more than image quality when the camera is static.

By Adrien Ramanana Rahary, Yutong Bai, Nicolas Dufour, Patrick P\'erez, David Picard
arXiv AI
Oct 31

Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion

Focused Forcing is a training‑free method for selecting key‑value (KV) cache entries in autoregressive video diffusion models. It selects historical frames per generated frame by combining attention and diversity scores, and allocates larger budgets to more important heads. The approach yields up to 1.48× faster generation while improving visual quality and text alignment across several generation paradigms.

By Peiliang Cai, Evelyn Zhang, Jiacheng Liu, Hao Lin, Ruiqi Zhang, Weile Mo, Yue Ma, Shikang Zheng, Jiehang Huang, Dongrui Liu, Linfeng Zhang
arXiv Computer Vision
Aug 31

Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generation

The paper introduces Relax Forcing, a training‑free memory mechanism for autoregressive video diffusion that structures temporal context into Sink, Tail, and History frames. By selecting History frames with a relaxation criterion, the method reduces error accumulation and attention overhead while preserving motion dynamics. Experiments on VBench‑Long demonstrate that this structured memory improves long‑video generation quality over existing baselines.

By Zengqun Zhao, Yanzuo Lu, Ziquan Liu, Jifei Song, Jiankang Deng, Ioannis Patras