What about gravity in video generation? Post-Training Newton's Laws with Verifiable Rewards
- URL: http://arxiv.org/abs/2512.00425v1
- Date: Sat, 29 Nov 2025 10:04:50 GMT
- Title: What about gravity in video generation? Post-Training Newton's Laws with Verifiable Rewards
- Authors: Minh-Quan Le, Yuanzhi Zhu, Vicky Kalogeiton, Dimitris Samaras,
- Abstract summary: Video diffusion models can synthesize visually compelling clips, yet often violate basic physical laws-objects float, accelerations drift, and collisions behave inconsistently-revealing a persistent gap between visual realism and physical realism.<n>We propose $textttNewtonRewards$, the first physics-grounded post-training framework for video generation based on $textitverifiable rewards$.
- Score: 49.02795965814016
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Recent video diffusion models can synthesize visually compelling clips, yet often violate basic physical laws-objects float, accelerations drift, and collisions behave inconsistently-revealing a persistent gap between visual realism and physical realism. We propose $\texttt{NewtonRewards}$, the first physics-grounded post-training framework for video generation based on $\textit{verifiable rewards}$. Instead of relying on human or VLM feedback, $\texttt{NewtonRewards}$ extracts $\textit{measurable proxies}$ from generated videos using frozen utility models: optical flow serves as a proxy for velocity, while high-level appearance features serve as a proxy for mass. These proxies enable explicit enforcement of Newtonian structure through two complementary rewards: a Newtonian kinematic constraint enforcing constant-acceleration dynamics, and a mass conservation reward preventing trivial, degenerate solutions. We evaluate $\texttt{NewtonRewards}$ on five Newtonian Motion Primitives (free fall, horizontal/parabolic throw, and ramp sliding down/up) using our newly constructed large-scale benchmark, $\texttt{NewtonBench-60K}$. Across all primitives in visual and physics metrics, $\texttt{NewtonRewards}$ consistently improves physical plausibility, motion smoothness, and temporal coherence over prior post-training methods. It further maintains strong performance under out-of-distribution shifts in height, speed, and friction. Our results show that physics-grounded verifiable rewards offer a scalable path toward physics-aware video generation.
Related papers
- GeoPT: Scaling Physics Simulation via Lifted Geometric Pre-Training [86.70824679370524]
We present GeoPT, a unified pre-trained model for general physics simulation based on lifted geometric pre-training.<n>The core idea is to augment geometry with synthetic dynamics, enabling dynamics-aware self-supervision without physics labels.
arXiv Detail & Related papers (2026-02-23T22:32:08Z) - PhysRVG: Physics-Aware Unified Reinforcement Learning for Video Generative Models [100.65199317765608]
Physical principles are fundamental to realistic visual simulation, but remain a significant oversight in transformer-based video generation.<n>We introduce a physics-aware reinforcement learning paradigm for video generation models that enforces physical collision rules directly in high-dimensional spaces.<n>We extend this paradigm to a unified framework, termed Mimicry-Discovery Cycle (MDcycle), which allows substantial fine-tuning.
arXiv Detail & Related papers (2026-01-16T08:40:10Z) - Objects in Generated Videos Are Slower Than They Appear: Models Suffer Sub-Earth Gravity and Don't Know Galileo's Principle...for now [10.272466104440381]
Video generators are increasingly evaluated as potential world models.<n>We investigate their representation of a fundamental law: gravity.<n>A lightweight low-rank adaptor fine-tuned on only 100 single-ball clips raises $g_mathrmeff$ from $1.81,mathrmm/s2$ to $6.43,mathrmm/s2$ (reaching $65% of terrestrial gravity).
arXiv Detail & Related papers (2025-12-01T18:59:56Z) - PhysCtrl: Generative Physics for Controllable and Physics-Grounded Video Generation [53.06495362038348]
Existing generation models excel at producing photo-realistic videos from text or images, but often lack physical plausibility and 3D controllability.<n>We introduce PhysCtrl, a novel framework for physics-grounded image-to-video generation with physical parameters and force control.<n> Experiments show that PhysCtrl generates realistic, physics-grounded motion trajectories which, when used to drive image-to-video models, yield high-fidelity, controllable videos.
arXiv Detail & Related papers (2025-09-24T17:58:04Z) - Physics-Grounded Motion Forecasting via Equation Discovery for Trajectory-Guided Image-to-Video Generation [54.42523027597904]
We introduce a novel framework that integrates symbolic regression and trajectory-guided image-to-video (I2V) models for physics-grounded video forecasting.<n>Our approach extracts motion trajectories from input videos, uses a retrieval-based pre-training mechanism to enhance symbolic regression, and discovers equations of motion to forecast physically accurate future trajectories.
arXiv Detail & Related papers (2025-07-09T13:28:42Z) - Force Prompting: Video Generation Models Can Learn and Generalize Physics-based Control Signals [18.86902152614664]
We investigate using physical forces as a control signal for video generation.<n>We propose force prompts which enable users to interact with images through both localized point forces.<n>We demonstrate that these force prompts can enable videos to respond realistically to physical control signals.
arXiv Detail & Related papers (2025-05-26T01:04:02Z) - MotionCraft: Physics-based Zero-Shot Video Generation [22.33113030344355]
MotionCraft is a new zero-shot video generator to craft physics-based and realistic videos.
We show that MotionCraft is able to warp the noise latent space of an image diffusion model, such as Stable Diffusion, by applying an optical flow.
We compare our method with the state-of-the-art Text2Video-Zero reporting qualitative and quantitative improvements.
arXiv Detail & Related papers (2024-05-22T11:44:57Z) - Machine-Learning Non-Conservative Dynamics for New-Physics Detection [69.45430691069974]
Given a trajectory governed by unknown forces, our Neural New-Physics Detector (NNPhD) aims to detect new physics.
We demonstrate that NNPhD successfully discovers new physics by decomposing the force field into conservative and non-conservative components.
We also show how NNPhD coupled with an integrator outperforms previous methods for predicting the future of a damped double pendulum.
arXiv Detail & Related papers (2021-05-31T18:00:10Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.