PulseAugur
EN
LIVE 07:49:11

New PhysEditWorld dataset enables physics-editable game world models

By PulseAugur Editorial · [2 sources] ·

Researchers have introduced PhysEditWorld, a large-scale dataset designed to enable physics-editable world models for game environments. This dataset focuses on gravity variations within 12 cinematic scenes rendered using Unreal Engine 5, capturing over 100 hours of gameplay and millions of frames. PhysEditWorld provides synchronized multimodal signals, including visual data, audio, action traces, and explicit gravity labels, facilitating research into controllable world modeling and improving the physical realism of generative video and world understanding models. AI

IMPACT Enables more physically realistic and controllable AI-generated game worlds.

RANK_REASON The cluster describes a new dataset released via arXiv, suitable for research purposes.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New PhysEditWorld dataset enables physics-editable game world models

COVERAGE [2]

  1. arXiv cs.CV TIER_1 English(EN) · Bin Hu, Yanwen Ma, Jiehui Huang, Ziliang Zhang, Haoning Wu, Ruicheng Zhang, Yaokun Li, Zijun Wang, Yuechen Zhang, Chun-Mei Tseng, Hanhui Li, Shengju Qian, Jun Zhou, Kaipeng Zhang, Xiaodan Liang, Jiaya Jia, Xiu Li ·

    PhysEditWorld: A Large-Scale Dataset Toward Physics-Editable World Models

    arXiv:2606.26694v1 Announce Type: new Abstract: Recent game world models can synthesize visually plausible, action-conditioned rollouts. However, their interaction behaviors often remain limited to exploratory or wandering trajectories, and physical dynamics are typically learned…

  2. arXiv cs.CV TIER_1 English(EN) · Xiu Li ·

    PhysEditWorld: A Large-Scale Dataset Toward Physics-Editable World Models

    Recent game world models can synthesize visually plausible, action-conditioned rollouts. However, their interaction behaviors often remain limited to exploratory or wandering trajectories, and physical dynamics are typically learned as implicit correlations from data rather than …

RELATED ENTITIES

RELATED TOPICS