VGGRPO: Towards World-Consistent Video Generation with 4D Latent Reward

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: An, Zhaochong, Kupyn, Orest, Uscidda, Théo, Colaco, Andrea, Ahuja, Karan, Belongie, Serge, Gonzalez-Franco, Mar, Gazulla, Marta Tintore
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912985280151552
author An, Zhaochong
Kupyn, Orest
Uscidda, Théo
Colaco, Andrea
Ahuja, Karan
Belongie, Serge
Gonzalez-Franco, Mar
Gazulla, Marta Tintore
author_facet An, Zhaochong
Kupyn, Orest
Uscidda, Théo
Colaco, Andrea
Ahuja, Karan
Belongie, Serge
Gonzalez-Franco, Mar
Gazulla, Marta Tintore
contents Large-scale video diffusion models achieve impressive visual quality, yet often fail to preserve geometric consistency. Prior approaches improve consistency either by augmenting the generator with additional modules or applying geometry-aware alignment. However, architectural modifications can compromise the generalization of internet-scale pretrained models, while existing alignment methods are limited to static scenes and rely on RGB-space rewards that require repeated VAE decoding, incurring substantial compute overhead and failing to generalize to highly dynamic real-world scenes. To preserve the pretrained capacity while improving geometric consistency, we propose VGGRPO (Visual Geometry GRPO), a latent geometry-guided framework for geometry-aware video post-training. VGGRPO introduces a Latent Geometry Model (LGM) that stitches video diffusion latents to geometry foundation models, enabling direct decoding of scene geometry from the latent space. By constructing LGM from a geometry model with 4D reconstruction capability, VGGRPO naturally extends to dynamic scenes, overcoming the static-scene limitations of prior methods. Building on this, we perform latent-space Group Relative Policy Optimization with two complementary rewards: a camera motion smoothness reward that penalizes jittery trajectories, and a geometry reprojection consistency reward that enforces cross-view geometric coherence. Experiments on both static and dynamic benchmarks show that VGGRPO improves camera stability, geometry consistency, and overall quality while eliminating costly VAE decoding, making latent-space geometry-guided reinforcement an efficient and flexible approach to world-consistent video generation.
format Preprint
id arxiv_https___arxiv_org_abs_2603_26599
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle VGGRPO: Towards World-Consistent Video Generation with 4D Latent Reward
An, Zhaochong
Kupyn, Orest
Uscidda, Théo
Colaco, Andrea
Ahuja, Karan
Belongie, Serge
Gonzalez-Franco, Mar
Gazulla, Marta Tintore
Computer Vision and Pattern Recognition
Large-scale video diffusion models achieve impressive visual quality, yet often fail to preserve geometric consistency. Prior approaches improve consistency either by augmenting the generator with additional modules or applying geometry-aware alignment. However, architectural modifications can compromise the generalization of internet-scale pretrained models, while existing alignment methods are limited to static scenes and rely on RGB-space rewards that require repeated VAE decoding, incurring substantial compute overhead and failing to generalize to highly dynamic real-world scenes. To preserve the pretrained capacity while improving geometric consistency, we propose VGGRPO (Visual Geometry GRPO), a latent geometry-guided framework for geometry-aware video post-training. VGGRPO introduces a Latent Geometry Model (LGM) that stitches video diffusion latents to geometry foundation models, enabling direct decoding of scene geometry from the latent space. By constructing LGM from a geometry model with 4D reconstruction capability, VGGRPO naturally extends to dynamic scenes, overcoming the static-scene limitations of prior methods. Building on this, we perform latent-space Group Relative Policy Optimization with two complementary rewards: a camera motion smoothness reward that penalizes jittery trajectories, and a geometry reprojection consistency reward that enforces cross-view geometric coherence. Experiments on both static and dynamic benchmarks show that VGGRPO improves camera stability, geometry consistency, and overall quality while eliminating costly VAE decoding, making latent-space geometry-guided reinforcement an efficient and flexible approach to world-consistent video generation.
title VGGRPO: Towards World-Consistent Video Generation with 4D Latent Reward
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.26599