AI Paper Atlas

Research map / Video / Video generation

Transform Trained Transformer for Accelerating Native 4K Video Generation

Jiangning Zhang, Junwei Zhu, Teng Hu, Yabiao Wang, Donghao Luo, Weijian Cao, Zhenye Gan, Xiaobin Hu, Zhucun Xue, Xiangtai Li, Chengjie Wang, Yong Liu

ICML 2026 · Video › Video generation · approach: Autoregressive long video generation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace T3 attention with shifted windows (reverse: use T3) (ablation)T3-Video-T2V-1.3B · 720P, 5K itersVQA: 67.34 -> 69.37Improved
use batch size 8 instead of 64 (ablation)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 66.65Hurt
increase batch size from 64 to 128 or 256 (ablation)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 68.72No clear effect
use large block ratio (3<ratio<6) layers only (ablation)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 67.14Hurt
use small block ratio (1<ratio<3) layers only (ablation)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 68.69Hurt
use lightweight deployment scheme (eVAE, 8 DiT steps)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 67.72Hurt
transform T3 back to full attention after training (ablation)T3-Video-T2V-1.3B · 720PVQA: 69.37 -> 69.51No clear effect
T3-Video vs UltraGen at 4KT3-Video-T2V-1.3B · 4K-VBenchVQA: 67.43 -> 71.72Improved
apply T3 to Wan2.2-T2V-5BWan2.2-5B · 4K-VBenchVQA: 47.23 -> 67.40Improved
apply T3 to Wan2.1-I2V-1.3BWan2.1-I2V-1.3B · 4K-VBenchVQA: 59.75 -> 63.6Improved
LoRA instead of full fine-tuning of T3T3-Video-T2V-1.3B · 4K-VBenchVQA: 71.72 -> 70.78Hurt
apply T3 attention at 4K for DiTWan2.1-T2V-1.3B · DiT latency at 2176x3840latency: 79,774.4 -> 4,165.8Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy