Saved in:
Bibliographic Details
Main Authors: Chen, Taiye, Ding, Zihan, Li, Anjian, Zhang, Christina, Xiao, Zeqi, Wang, Yisen, Jin, Chi
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.12940
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912713331965952
author Chen, Taiye
Ding, Zihan
Li, Anjian
Zhang, Christina
Xiao, Zeqi
Wang, Yisen
Jin, Chi
author_facet Chen, Taiye
Ding, Zihan
Li, Anjian
Zhang, Christina
Xiao, Zeqi
Wang, Yisen
Jin, Chi
contents Recent advancements in video generation have demonstrated the potential of using video diffusion models as world models, with autoregressive generation of infinitely long videos through masked conditioning. However, such models, usually with local full attention, lack effective memory compression and retrieval for long-term generation beyond the window size, leading to issues of forgetting and spatiotemporal inconsistencies. To enhance the retention of historical information within a fixed memory budget, we introduce a recurrent neural network (RNN) into the diffusion transformer framework. Specifically, a diffusion model incorporating LSTM with attention achieves comparable performance to state-of-the-art RNN blocks, such as TTT and Mamba2. Moreover, existing diffusion-RNN approaches often suffer from performance degradation due to training-inference gap or the lack of overlap across windows. To address these limitations, we propose a novel Recurrent Autoregressive Diffusion (RAD) framework, which executes frame-wise autoregression for memory update and retrieval, consistently across training and inference time. Experiments on Memory Maze and Minecraft datasets demonstrate the superiority of RAD for long video generation, highlighting the efficiency of LSTM in sequence modeling.
format Preprint
id arxiv_https___arxiv_org_abs_2511_12940
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Recurrent Autoregressive Diffusion: Global Memory Meets Local Attention
Chen, Taiye
Ding, Zihan
Li, Anjian
Zhang, Christina
Xiao, Zeqi
Wang, Yisen
Jin, Chi
Computer Vision and Pattern Recognition
Recent advancements in video generation have demonstrated the potential of using video diffusion models as world models, with autoregressive generation of infinitely long videos through masked conditioning. However, such models, usually with local full attention, lack effective memory compression and retrieval for long-term generation beyond the window size, leading to issues of forgetting and spatiotemporal inconsistencies. To enhance the retention of historical information within a fixed memory budget, we introduce a recurrent neural network (RNN) into the diffusion transformer framework. Specifically, a diffusion model incorporating LSTM with attention achieves comparable performance to state-of-the-art RNN blocks, such as TTT and Mamba2. Moreover, existing diffusion-RNN approaches often suffer from performance degradation due to training-inference gap or the lack of overlap across windows. To address these limitations, we propose a novel Recurrent Autoregressive Diffusion (RAD) framework, which executes frame-wise autoregression for memory update and retrieval, consistently across training and inference time. Experiments on Memory Maze and Minecraft datasets demonstrate the superiority of RAD for long video generation, highlighting the efficiency of LSTM in sequence modeling.
title Recurrent Autoregressive Diffusion: Global Memory Meets Local Attention
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.12940