AI Paper Atlas

Research map / Video / Video generation

How Far are AI-generated Videos from Simulating the 3D Visual World: A Learned 3D Evaluation Approach

Chirui Chang, Jiahui Liu, Zhengzhe Liu, Xiaoyang Lyu, Yi-Hua Huang, Xin Tao, Pengfei Wan, Di Zhang, Xiaojuan Qi

ICCV 2025 · Video › Video generation · approach: World-model video generation · 29 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 contrastive loss (ablation)L3DE with concatenation fusion · Real vs synthetic detectionaccuracy (%): 68.77 -> 73.14Improved
add contrastive loss (ablation)L3DE with element-wise addition fusion · Real vs synthetic detectionaccuracy (%): 66.01 -> 69.26Improved
use concatenation instead of element-wise addition (ablation)L3DE without contrastive loss · Real vs synthetic detectionaccuracy (%): 66.01 -> 68.77Improved
fuse appearance, motion and geometry instead of appearance only (ablation)L3DE · Correlation with 3D reconstruction qualitySpearman correlation: 0.7181 -> 0.7566Improved
fuse proxies instead of appearance only (ablation)L3DE · Correlation with human ratingsSpearman correlation: 0.5643 -> 0.6460Improved
use geometry proxy only instead of fusion (ablation)L3DE · Correlation with 3D reconstruction qualitySpearman correlation: 0.7566 -> 0.3142Hurt
train on static plus dynamic scenes instead of static only (ablation)L3DE · Mixed-motion test setaccuracy (%): 69.55 -> 77.70Improved
train on indoor plus outdoor instead of indoor only (ablation)L3DE · Mixed indoor-outdoor test setaccuracy (%): 67.60 -> 76.55Improved
use L3DE instead of VBench background consistencyhuman rating correlation · 300 rated videosSpearman's rho: 20.68 -> 64.84Improved
use L3DE instead of EvalCrafterhuman annotation correlation · ECTV visual qualitycorrelation (%): 55.4 -> 67.0Improved
use L3DE instead of EvalCrafterhuman annotation correlation · ECTV motion qualitycorrelation (%): 45.0 -> 43.6No clear effect
use L3DE instead of NPR for fake video detectionfake video detection · Fake detection averageaccuracy (%): 54.25 -> 73.14Improved

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