PhysRVG: Physics-Aware Unified Reinforcement Learning for Video Generative Models

2026-01-16

AI Summary

This work addresses the lack of physical plausibility in existing Transformer-based video generation models, which often disregard rigid-body physics during pixel-level denoising, leading to unrealistic behaviors in collision scenarios. To overcome this limitation, we propose a physics-aware reinforcement learning paradigm that, for the first time, explicitly embeds Newtonian mechanics–driven collision rules as reinforcement signals directly into the high-dimensional generative space, rather than imposing them as post-hoc constraints. We introduce the Mimicry-Discovery Cycle (MDcycle), a unified framework that preserves physical feedback during large-scale fine-tuning, enabling co-optimization of physical fidelity and generative flexibility. Experiments on the newly established PhysRVGBench benchmark demonstrate that our approach significantly outperforms current methods in both physical realism and rigid-body motion consistency.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Physical principles are fundamental to realistic visual simulation, but remain a significant oversight in transformer-based video generation. This gap highlights a critical limitation in rendering rigid body motion, a core tenet of classical mechanics. While computer graphics and physics-based simulators can easily model such collisions using Newton formulas, modern pretrain-finetune paradigms discard the concept of object rigidity during pixel-level global denoising. Even perfectly correct mathematical constraints are treated as suboptimal solutions (i.e., conditions) during model optimization in post-training, fundamentally limiting the physical realism of generated videos. Motivated by these considerations, we introduce, for the first time, a physics-aware reinforcement learning paradigm for video generation models that enforces physical collision rules directly in high-dimensional spaces, ensuring the physics knowledge is strictly applied rather than treated as conditions. Subsequently, we extend this paradigm to a unified framework, termed Mimicry-Discovery Cycle (MDcycle), which allows substantial fine-tuning while fully preserving the model's ability to leverage physics-grounded feedback. To validate our approach, we construct new benchmark PhysRVGBench and perform extensive qualitative and quantitative experiments to thoroughly assess its effectiveness.

Problem

• physics-aware
• video generation
• rigid body motion
• physical realism
• collision modeling

Innovation

• physics-aware reinforcement learning
• video generative models
• rigid body motion
• Mimicry-Discovery Cycle
• physical realism