AI Paper Atlas

Research map / Video / Video generation

DeRA: Decoupled Representation Alignment for Video Tokenization

Pengbo Guo, Junke Wang, Zhen Xing, Chengxu Liu, Daoguo Dong, Xueming Qian, Zuxuan Wu

ECCV 2026 · Video › Video generation · approach: Compositional text-to-video diffusion · 5 citations (OpenAlex)

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
add DINOv3 alignment to appearance stream (ablation)DeRA base tokenizer · UCF-101rFVD: 24.47 -> 20.14Improved
add InternVideo2 alignment to motion stream (ablation)DeRA base tokenizer · UCF-101rFVD: 24.47 -> 22.64Improved
align both streams with separate foundation models (ablation)DeRA base tokenizer · UCF-101rFVD: 24.47 -> 18.83Improved
align both streams (ablation)DeRA base tokenizer · UCF-101gFVD: 112 -> 94Improved
use DINOv3 instead of DINOv2 as image target (ablation)DeRA · UCF-101rFVD: 23.98 -> 20.14Improved
use InternVideo2 instead of VideoMAEv2 as video target (ablation)DeRA · UCF-101rFVD: 24.14 -> 22.64Improved
add SACP gradient reformulation (ablation)DeRA · UCF-101rFVD: 19.81 -> 18.83Improved
add soft gradient-conflict loss with weight 0.5 (ablation)DeRA · UCF-101gFVD: 97 -> 96Improved
add soft gradient-conflict loss with weight 1 (ablation)DeRA · UCF-101gFVD: 97 -> 103Hurt
two-stage alignment instead of joint one-stage (ablation)DeRA · UCF-101rFVD: 18.83 -> 19.52No clear effect
raise lambda_m from 0.5 to 1 (ablation)DeRA · UCF-101rFVD: 18.83 -> 18.94No clear effect
align deeper encoder layer (depth 6 to 12) with DINOv3 (ablation)DeRA · UCF-101rFVD: 21.34 -> 20.14Improved

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