Improving the Physics of Video Generation with VJEPA-2 Reward Signal

TL;DR

This paper demonstrates that integrating the VJEPA-2 reward signal with a state-of-the-art video generative model enhances its physics plausibility by approximately 6%, addressing the gap between visual realism and physical understanding.

Contribution

The study introduces a novel approach of using VJEPA-2 as a reward signal to improve physics plausibility in video generation models, building on SSL-based world models.

Findings

Physics plausibility improved by ~6% using VJEPA-2

Leveraging SSL-based models enhances physical understanding in video generation

VJEPA-2 effectively guides generative models towards more physically consistent videos

Abstract

This is a short technical report describing the winning entry of the PhysicsIQ Challenge, presented at the Perception Test Workshop at ICCV 2025. State-of-the-art video generative models exhibit severely limited physical understanding, and often produce implausible videos. The Physics IQ benchmark has shown that visual realism does not imply physics understanding. Yet, intuitive physics understanding has shown to emerge from SSL pretraining on natural videos. In this report, we investigate whether we can leverage SSL-based video world models to improve the physics plausibility of video generative models. In particular, we build ontop of the state-of-the-art video generative model MAGI-1 and couple it with the recently introduced Video Joint Embedding Predictive Architecture 2 (VJEPA-2) to guide the generation process. We show that by leveraging VJEPA-2 as reward signal, we can improve the physics plausibility of state-of-the-art video generative models by ~6%.