SOURCE-LINKED INTELLIGENCE

CAER: Causal Action Effect Reweighting for World Model Training

arXiv · AI, language, vision and robotics · article · Aug 31, 2026 · UTC

World models are becoming core infrastructure for embodied intelligence, with action-conditioned video generation providing controllable predictions of how scenes evolve after agent interventions. Yet existing models are commonly trained with space-time-uniform mean squared error, allowing abundant background tokens to dominate the gradient while sparse interaction dynamics remain under-optimized; such uniform fitting rewards reconstructing appearance rather than learning how actions change the world. We introduce Causal Action Effect Reweighting (CAER), a general training paradigm that redist

Read original source ↗ Open in workspace

recordType

paper

region

Global