The paper introduces Connected Self Forcing, a training framework that reconnects gradient paths across autoregressive video chunks to improve long‑horizon visual quality and temporal consistency. It allows feedback from later predictions to influence earlier context generation, using shortcut gradient replay to keep memory usage manageable. The method is combined with distribution‑matching distillation and shows experimental gains without altering the inference procedure.
By Dongbin Zhang, Chaoda Zheng, Kangjie Chen, Xiangyu Li, Shijia Chen, Jinhao Deng, Yuqi Zhang, Guangfeng Jiang, Hongbin Lin, Choo Sin Wai, Minqi Wang, Puyi Wang, Jingye Zhang, Yu Zhang, Xianming Liu, Boyang Wang
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
arXiv:2609.37925v1 Announce Type: cross
Abstract: Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Trai...
By Chenjian Gao, Zhihao Hu, Jianqi Ma, Jun Zhang, Weidong Zhang, Tianfan Xue
arXiv:2512.21004v2 Announce Type: replace
Abstract: Recent advances in pretraining general foundation models have significantly improved performance across diverse downstream tasks. While autoregress...
By Jinghan Li, Yang Jin, Hao Jiang, Yadong Mu, Yang Song, Kun Xu
arXiv:2608. 05237v1 Announce Type: cross Abstract: Current few-step autoregressive video diffusion models depend on previous fully denoised clean frames as context for all denoising steps of the current frame.
By Lingxiao Yang, Liu Liu, Moran Li, Han Feng, Wenjian Cao, Jiangning Zhang, Ye Shi
arXiv:2607. 06481v1 Announce Type: cross Abstract: We present PACR-Video, a parameter-efficient framework for multi-shot long video extrapolation that preserves recurring entities, scene structure, visual style, and causal progression without full generator fine-tuning.
By Anna C\'ordoba, Adam Puente Tercero, Nerea Angulo Hijo, Mar Linares Tercero, Julia Barrientos, Ainhoa Miranda, Jes\'us Olivera
arXiv:2610.13179v1 Announce Type: new
Abstract: Generative video super-resolution (VSR) can recover visually compelling high-frequency detail by leveraging diffusion priors, but iterative sampling an...
By Georgios Panaretou, Sotirios Chatzis
The paper introduces Prediction‑Aligned Context Compaction (PACC), a method that learns a compact memory representation for long‑video generation by distilling a frozen video generator. PACC trains a compressor to aggregate past frames into memory tokens, using the generator as both teacher and student during on‑policy distillation. Experiments on MBench and VBench‑Long show that PACC improves memory‑event coverage and consistency, achieving better scores than strong baselines and producing competitive minute‑long videos.
By Xiaoyu Wu, Weihang Guo, Yifei Wang, Xinze Feng, Lydia E. Kavraki, Zhiwei Steven Wu
Latent video generation relies on autoencoders to define a compact space in which generative models operate. Although video autoencoder architectures have evolved substantially, their latent spaces are still optimized primarily for pixel-level reconstruction and provide limited high-level semantic organization.
LongVU‑TTT is a causal test‑time training method for long‑video multimodal large language models that inserts a convolutional resampler with fast‑weight updates between the vision encoder and the LLM. The fast weights adapt per video and contextualize frame features before compression, while a hybrid selector keeps explicit visual evidence for downstream reasoning. Experiments show that TTT‑Conv outperforms TTT‑MLP and bidirectional Mamba2 on MLVU, and beats attention‑ and fixed‑state recurrent resamplers on three benchmarks, achieving competitive results on five video‑understanding tasks after reducing 512 frames to 128 LLM frames.
By Mahmoud Ahmed, Sameh Abdulah, Olatunji Ruwase, Sam Ade Jacobs, Mathis Bode, Mohamed Elhoseiny
arXiv:2610.02914v1 Announce Type: new
Abstract: Autoregressive video models can generate minute-long videos in real time, but they produce generic subjects from text rather than specific subjects fro...
By Yunseung Ok (Kyung Hee University), Hyunsoo Kim (The University of Texas at Austin), Minseo Kim (Kyung Hee University), Suhyun Kim (Kyung Hee University)
High-Fidelity Video Quality Assessment (HFVQA) is a new framework that uses fixed-size spatio‑temporal patches across multiple scales, including the original resolution, to preserve low‑level quality cues and semantic context. It incorporates a lightweight auxiliary network that learns VQA‑specific saliency directly from quality supervision, enabling the model to focus on the most important spatio‑temporal regions. By combining high‑fidelity cues with task‑specific saliency, HFVQA achieves state‑of‑the‑art performance on standard no‑reference VQA benchmarks while processing only about 12% of the candidate patches, making it computationally efficient.
By Hakan Emre Gedik, Shashank Gupta, Alan Bovik