Innovation
Methods, ideas, or system contributions that make the work stand out.
Embodied World Models
Physical Evolution Supervision
Instance-Guided Restoration
Temporal Instance Alignment
Model Laziness Rate
EVEWorld: Physical Evolution Supervision for Embodied World Models
🤖 AI Summary
This study addresses the issue of "model inertia" in embodied world models arising from deficient physical reasoning by proposing a physics evolution supervision framework. Methodologically, it introduces a novel evaluation metric termed the model inertia rate and designs a dual-component process-level supervision mechanism comprising Instance-Guided Recovery (IGR) and Temporal Instance Alignment (TIA). These components collectively ensure temporal dynamic physical consistency and plausibility throughout the generation process. Experimental results demonstrate that the proposed framework achieves superior performance across multiple benchmarks, reducing the model inertia rate by 87.5% compared to GigaWorld-0. Consequently, this work effectively enhances the physical reasoning capabilities of embodied world models.
📝 Abstract
Embodied world models enable scalable simulation of embodied interactions for robot learning. However, existing models are prone to Model Laziness, as they focus on visual fidelity at the expense of physical reasoning and lack process-level supervision over the temporal dynamics of manipulated objects. In this work, we propose EVEWorld, a physical evolution-supervision framework for physically consistent target evolution. EVEWorld consists of two components: Instance-Guided Restoration (IGR) and Temporal Instance Alignment (TIA). First, IGR promotes instance consistency through restoration supervision. Second, TIA promotes cross-frame consistency by aligning target instances across adjacent frames. We further introduce the Model Laziness Rate (MLR), a metric that measures persistent violations of instance consistency in generated trajectories. Extensive experiments on DreamGenBench, EWMBench, and PBench demonstrate the effectiveness of EVEWorld, notably achieving an 87.5% reduction in MLR compared with GigaWorld-0. On the WorldArena 2.0 Track 1 leaderboard, our model ranks 6th in JEPA Similarity and 17th overall, which further validates the performance of our evolution supervision strategy.
🔎 Similar Papers
No similar papers found.
