AI Paper Atlas

Research map / Video / Video-language understanding

DynaVieW: Schema-Guided World Modeling for Understanding Hierarchical Visual Dynamics

Silin Gao, Hao Zhao, Zeming Chen, Sepideh Mamooler, Antara Raaghavi Bhattacharya, Qiyu Wu, Hiromi Wakaki, Yuki Mitsufuji, Li Mi, Syrielle Montariol, Antoine Bosselut

ICML 2026 · Video › Video-language understanding · approach: Event and anomaly reasoning in video · 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
continue pre-training BAGEL on hierarchical state-transition data (DynaVieW) before zero-shot narrative generationBAGEL-7B-MoT · VinaBench VWP zero-shotaverage score: 0.567 -> 0.630Improved
continue pre-training on state-transition data then SFT on VinaBenchBAGEL-7B-MoT · VinaBench VWP SFTaverage score: 0.547 -> 0.610Improved
add SFT on task data to DynaVieWDynaVieW · VinaBench VWPcross-scene consistency (character): 0.550 -> 0.483Mixed
add multimodal selective attention (ablation)DynaVieW · VinaBench VWP zero-shotaverage score: -0.013 when removedImproved
add schema token re-weighted CE loss (ablation)DynaVieW · VinaBench VWP zero-shotaverage score: -0.004 when removedImproved
use JSON-schema transitions instead of template natural language (ablation)DynaVieW training data · VinaBench VWP zero-shotaverage score: 0.028 average decrease when removedImproved
use fine-grained low-level transitions instead of coarse schema (ablation)DynaVieW training data · VinaBench VWP zero-shotaverage score: 0.064 average decrease when removedImproved
use DynaVieW instead of BAGEL with LLM-generated JSON schema promptsnext-scene control · VinaBench controllabilityaverage score: 0.283 -> 0.604Improved
convert JSON schema into natural language for BAGEL (ablation)BAGEL · VinaBench controllabilityaverage score: 0.283 -> 0.275No clear effect
use DynaVieW instead of BAGEL for world simulation, no task SFTBAGEL-7B-MoT · LEGO Ego4DFID: 26.94 -> 26.15Improved
use DynaVieW instead of BAGEL for world simulation after SFTBAGEL-7B-MoT · LEGO Epic-Kitchens (SFT)FID: 11.66 -> 10.31Improved
use DynaVieW instead of BAGEL (zero-shot)BAGEL-7B-MoT · LEGO Epic-KitchensLPIPS: 41.45 -> 41.78Mixed

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