AI Paper Atlas

Research map / Video / Action recognition

Olaf-World: Orienting Latent Actions for Video World Modeling

Yuxin Jiang, Yuchao Gu, Ivor Tsang, Mike Zheng Shou

ICML 2026 · Video › Action recognition · approach: Action anticipation and latent actions · 0 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 Seq-Delta-REPA effect alignment to latent action learningAdaWorld-style beta-VAE LAM · MIND 1st-P to 1st-P linear probeMacro-F1: 0.6004 -> 0.8138Improved
add Seq-Delta-REPA effect alignmentAdaWorld-style beta-VAE LAM · MIND 3rd-P to 3rd-P linear probeMacro-F1: 0.4827 -> 0.8256Improved
add Seq-Delta-REPA effect alignmentAdaWorld-style beta-VAE LAM · MIND cross-domain probe 1st-P to 3rd-PMacro-F1: 0.4820 -> 0.6250Improved
add Seq-Delta-REPA effect alignment, adapt with 50 videosOlaf-World vs AdaWorld · MIND 3RD-PRPE rotation: 0.3353 -> 0.2082Improved
add Seq-Delta-REPA, 1 adaptation videoOlaf-World vs AdaWorld · MIND 1ST-PRPE rotation: 0.6420 -> 0.4680Improved
add Seq-Delta-REPA effect alignment, video qualityOlaf-World vs AdaWorld · MIND 1ST-P, 0 videosImage quality: 0.5600 -> 0.5400Hurt
latent-action pretraining vs direct ground-truth action conditioningOlaf-World vs DirectAct · MIND 1ST-P, 1 videoRPE rotation: 1.2822 -> 0.4680Improved
add latent-action pretraining, unseen visual contextsOlaf-World vs AdaWorld · OOD 50 initial framesRPE rotation: 1.7063 -> 1.2221Improved
align temporal feature differences instead of static features (ablation)Seq-Delta-REPA · MIND 1st-P to 3rd-P probeMacro-F1: 0.5287 -> 0.6250Improved
use L2-normalized cosine alignment instead of MSE (ablation)Seq-Delta-REPA · MIND 3rd-P to 3rd-P probeMacro-F1: 0.7096 -> 0.8256Improved
use video reference encoder (V-JEPA2) instead of image encoder (DINOv3) (ablation)Seq-Delta-REPA · MIND 1st-P to 1st-P probeMacro-F1: 0.2334 -> 0.7044Improved
use DINOv3 image encoder as reference vs no alignment (ablation)Seq-Delta-REPA · MIND 1st-P to 1st-P probeMacro-F1: 0.1849 -> 0.2334No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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