PhysForge: Generating Physics-Grounded 3D Assets for Interactive Virtual Worlds

Forum topic · 小凯 · 2026-05-08

Summary

PhysForge is a two-stage framework for generating simulation-ready, physics-grounded 3D assets, addressing a key bottleneck for interactive virtual worlds and embodied AI. Unlike existing methods that focus only on static geometry, PhysForge roots asset generation in functional logic and hierarchical physics. It is supported by PhysDB, a large-scale dataset of 150,000 assets with four-tier physical annotations. In the first stage, a vision-language model acts as a physical architect to plan a Hierarchical Physical Blueprint covering material, functional, and kinematic constraints. In the second stage, a physics-grounded diffusion model realizes the blueprint by synthesizing high-fidelity geometry together with precise kinematic parameters using a novel KineVoxel Injection (KVI) mechanism. Experiments show that PhysForge produces functionally plausible, simulation-ready assets, serving as a robust data engine for interactive 3D content and embodied agents. Paper: arXiv 2605.05163.

Paper Overview

Field: Computer Vision (CV)

Authors: Yunhan Yang, Chunshi Wang, Junliang Ye, Yang Li, Zanxin Chen, Zehuan Huang, Yao Mu, Zhuo Chen, Chunchao Guo, Xihui Liu

Published: 2026-05-06

arXiv: 2605.05163

Abstract

Synthesizing physics-grounded 3D assets is a critical bottleneck for interactive virtual worlds and embodied AI. Existing methods predominantly focus on static geometry, overlooking the functional properties essential for interaction. The authors propose that interactive asset generation must be rooted in functional logic and hierarchical physics.

To bridge this gap, PhysForge is introduced: a decoupled two-stage framework supported by PhysDB, a large-scale dataset of 150,000 assets with four-tier physical annotations.

1. Stage 1 — Physical planning: A VLM acts as a physical architect to plan a Hierarchical Physical Blueprint defining material, functional, and kinematic constraints. 2. Stage 2 — Physics-grounded synthesis: A physics-grounded diffusion model realizes this blueprint by synthesizing high-fidelity geometry alongside precise kinematic parameters via a novel KineVoxel Injection (KVI) mechanism.

Experiments demonstrate that PhysForge produces functionally plausible, simulation-ready assets, providing a robust data engine for interactive 3D content and embodied agents.

Key Contributions

A decoupled two-stage pipeline separating physical reasoning (VLM planning) from generation (diffusion-based synthesis)
PhysDB: 150,000 assets annotated with four tiers of physical information
KineVoxel Injection (KVI): a mechanism for injecting kinematic parameters into geometry synthesis
Simulation-ready assets that support interactive 3D content and embodied AI

---

*Source: forum post, auto-collected on 2026-05-08.*