MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse

Papers with Code Trending Papers

Summary

MetaSpatial is a reinforcement learning framework that enhances 3D spatial reasoning in vision-language models, enabling coherent and physically plausible 3D scene generation without hard-coded optimizations.

We present MetaSpatial, the first reinforcement learning (RL)-based framework designed to enhance 3D spatial reasoning in vision-language models (VLMs), enabling real-time 3D scene generation without the need for hard-coded optimizations. MetaSpatial addresses two core challenges: (i) the lack of internalized 3D spatial reasoning in VLMs, which limits their ability to generate realistic layouts, and (ii) the inefficiency of traditional supervised fine-tuning (SFT) for layout generation tasks, as perfect ground truth annotations are unavailable. Our key innovation is a multi-turn RL-based optimization mechanism that integrates physics-aware constraints and rendered image evaluations, ensuring generated 3D layouts are coherent, physically plausible, and aesthetically consistent. Methodologically, MetaSpatial introduces an adaptive, iterative reasoning process, where the VLM refines spatial arrangements over multiple turns by analyzing rendered outputs, improving scene coherence progressively. Empirical evaluations demonstrate that MetaSpatial significantly enhances the spatial consistency and formatting stability of various scale models. Post-training, object placements are more realistic, aligned, and functionally coherent, validating the effectiveness of RL for 3D spatial reasoning in metaverse, AR/VR, digital twins, and game development applications. Our code, data, and training pipeline are publicly available at https://github.com/PzySeere/MetaSpatial.
Original Article
View Cached Full Text

Cached at: 07/05/26, 10:32 AM

Paper page - MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse

Source: https://huggingface.co/papers/2503.18470 Published on Mar 24, 2025

Abstract

MetaSpatial is a reinforcement learning framework that enhances 3D spatial reasoning in vision-language models, enabling realistic and coherent 3D scene generation through iterative refinement and physics-aware constraints.

We present MetaSpatial, the firstreinforcement learning(RL)-based framework designed to enhance3D spatial reasoninginvision-language models(VLMs), enablingreal-time 3D scene generationwithout the need for hard-coded optimizations. MetaSpatial addresses two core challenges: (i) the lack of internalized3D spatial reasoningin VLMs, which limits their ability to generate realistic layouts, and (ii) the inefficiency of traditional supervised fine-tuning (SFT) for layout generation tasks, as perfect ground truth annotations are unavailable. Our key innovation is amulti-turn RL-based optimization mechanism that integratesphysics-aware constraintsand rendered image evaluations, ensuring generated 3D layouts are coherent, physically plausible, and aesthetically consistent. Methodologically, MetaSpatial introduces an adaptive, iterative reasoning process, where the VLM refines spatial arrangements over multiple turns by analyzing rendered outputs, improving scene coherence progressively.Empirical evaluationsdemonstrate that MetaSpatial significantly enhances the spatial consistency and formatting stability of various scale models. Post-training, object placements are more realistic, aligned, and functionally coherent, validating the effectiveness of RL for3D spatial reasoninginmetaverse,AR/VR,digital twins, and game development applications. Our code, data, and training pipeline are publicly available at https://github.com/PzySeere/MetaSpatial.

View arXiv pageView PDFProject pageGitHub322Add to collection

Get this paper in your agent:

hf papers read 2503\.18470

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2503.18470 in a model README.md to link it from this page.

Datasets citing this paper3

#### johnschaefer/EasyR1-qwen3vl-rl Updated22 days ago • 577 #### Yuting6/ttrl Viewer• UpdatedApr 30, 2025 • 6 • 434 #### zhenyupan/3d_layout_reasoning Viewer• UpdatedMar 25, 2025 • 100 • 117

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2503.18470 in a Space README.md to link it from this page.

Collections including this paper2

Similar Articles

SpatialAct: Probing Spatial Reasoning-to-Action Capabilities of VLM Agents in 3D Scenes

Hugging Face Daily Papers

SpatialAct is a new simulator-grounded benchmark that probes whether VLM agents can perform coherent spatial reasoning and translate it into actions in 3D environments across multi-turn feedback settings. Experiments reveal a significant reasoning-to-action gap, with current VLMs struggling to maintain spatial beliefs and produce reliable actions despite performing well on isolated reasoning tasks.

Reinforcing Dual-Path Reasoning in Spatial Vision Language Models

Hugging Face Daily Papers

This paper introduces SR-REAL, a unified framework for spatial vision-language models that combines linguistic deduction and 3D geometric reasoning via reinforcement learning, enabling robust multi-step spatial reasoning across diverse tasks.