Generation Models Know Space: Unleashing Implicit 3D Priors for Scene Understanding

Published June 30, 2026 · Category: Robotics

Overview

arXiv:2603.19235v2 Announce Type: replace-cross Abstract: While Multimodal Large Language Models demonstrate impressive semantic capabilities, they often suffer from spatial blindness, struggling with fine-grained geometric reasoning and physical dynamics. Existing solutions typically rely on explicit 3D modalities or complex geometric scaffolding, which are limited by data scarcity and generalization challenges. In this work, we propose a paradigm shift by leveraging the implicit spatial prior within large-scale video generation models. We posit that to synthesize temporally coherent videos, these models inherently learn robust 3D structural priors and physical laws. We introduce VEGA-3D (Video Extracted Generative Awareness), a plug-and-play framework that repurposes a pre-trained video diffusion model as a Latent World Simulator. By extracting spatiotemporal features from intermediate noise levels and integrating them with semantic representations via a token-level adaptive gated fusion mechanism, we enrich MLLMs with dense geometric cues without explicit 3D supervision. Extensive experiments across 3D scene understanding, spatial reasoning, and embodied manipulation benchmarks demonstrate that our method outperforms state-of-the-art baselines, validating that generative priors provide a scalable foundation for physical-world understanding. Code is publicly available at https://github.com/H-EmbodVis/VEGA-3D.

Source

Originally published at arxiv.org.

Related Articles

• You Only Touch Once: 6-DoF Object Pose Estimation from Single Tactile Contact
• The Speedup Paradox: Rethinking Inference Speed-Quality Trade-off in Embodied Tasks
• X-Morph: Human Motion Priors for Scalable Robot Learning Across Morphologies