AI Paper Atlas

Research map / Video / Video generation

Co-Evolving Latent Action World Models

Yucen Wang, Fengming Zhang, De-Chuan Zhan, Li Zhao, Kaixin Wang, Jiang Bian

ICML 2026 · Video › Video generation · approach: World-model 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
add warm-up phase (frozen world model supplies gradients to IDM) before joint training (ablation)CoLA-World joint IDM + OpenSora training · latent action codebook metricscodebook utilization/entropyImproved
joint training without warm-up (random or pretrained IDM init) (ablation)IDM + pretrained OpenSora world model · codebook metricscodebook utilizationHurt
longer warm-up before joint training (ablation)CoLA-World · codebook metricsjoint training stabilityImproved
joint training instead of two-stage with same 60K budgetOpenSora world model · OXE video predictionFVD: 291.30 -> 278.90Improved
joint training instead of two-stage on real-action adaptationOpenSora world model · LIBERO real actionFVD: 115.45 -> 93.68Improved
joint training instead of two-stage on visual planningOpenSora world model · VP2 RoboDesk (5 tasks)average success rate: 7.73% -> 21.20%Improved
keep LAM frozen after warm-up instead of co-evolving with world model (ablation)CoLA-World · world model video predictionvideo generation qualityHurt
continue E2E training with unfrozen world model instead of pure warm-up (frozen world model) (ablation)CoLA-World LAM · LIBERO linear probingprobing lossImproved
joint training instead of two-stage for LAM qualitylinear probing of latent actions · six robotics datasetsL1 probing loss: 0.1614 -> 0.1623 on LIBERONo clear effect
scale world model size (6 to 28 DiT blocks) (ablation)CoLA-World · LIBERO linear probingprobing lossImproved
increase batch size from 64 to 128 (ablation)CoLA-World · video prediction and LIBERO probingFVD: 261.78 -> 233.06Improved
swap OpenSora backbone for Wan2.1 (ablation)CoLA-World · OXE, EgoCentric, AgiBot, LIBEROFVD: 169.82 -> 155.58 on OXE; 161.91 -> 168.96 on LIBEROMixed

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