GEOPHYS: The Geometry of Physical Plausibility
Provides a fast, training-free method for physics-violation detection and video generation verification, addressing the need for efficient physical plausibility assessment in video understanding and generation.
GEOPHYS uses five geometric properties of frozen image encoder embeddings to detect physical implausibility in videos, achieving 98.3% on LikePhys and 93.3% on IntPhys2, outperforming GPT-4o, Gemini, and video diffusion models. As a verifier for video generation, it improves MAGI-1 24B from 50.01% to 64.50% on PhysicsIQ with lower compute and memory.
While humans can identify physically implausible events within milliseconds, machine learning approaches addressing the same problem are extremely slow and expensive. They either rely on external multimodal-LLM judges or require ad-hoc modifications to the training procedure. In this work, we argue that indicators of physical plausibility are implicitly captured by five geometric properties of the per-frame embeddings produced by frozen image encoders. In aggregate, we call them GEOPHYS. First, we show that these signals correlate with human EEG responses to two forms of object-permanence violations. Second, GEOPHYS robustly discriminates physically implausible videos from realistic ones, achieving state-of-the-art physics-violation detection: 98.3% on LikePhys and 93.3% on IntPhys2, whereas V-JEPA 2, GPT-4o, Gemini, and twelve modern video diffusion models perform near chance. Third, used as a best-of-N verifier for physical alignment during video generation, GEOPHYS lifts MAGI-1 24B from 50.01% to 64.50% on PhysicsIQ at 1.5x lower wall-clock and 4.65x lower memory than the V-JEPA 2 world-model verifier. Ultimately, GEOPHYS demonstrates that physical plausibility in videos can be assessed by leveraging the emergent geometric properties of temporal features extracted from image encoders.