Research map / Probabilistic and scientific modeling / Dynamical systems
UniPixie: Unified and Probabilistic 3D Physics Learning via Flow Matching
Qilin Huang, Quynh Anh Huynh, Long Le, Chen Wang, Chuhao Chen, Ryan Lucas, Eric Eaton, Lingjie Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace 3D U-Net diffusion with flow matching transformer (ablation) | UniPixie · PixieMultiVerse test | log E MSE: 0.0410 -> 0.0091 | Improved |
| replace 3D U-Net diffusion with flow matching transformer (ablation) | UniPixie · PixieMultiVerse test | log rho MSE: 0.1464 -> 0.0194 | Improved |
| replace 3D U-Net diffusion with flow matching transformer (ablation) | UniPixie · PixieMultiVerse test | nu MSE: 0.4604 -> 0.0240 | Improved |
| replace 3D U-Net diffusion with flow matching transformer (ablation) | UniPixie · PixieMultiVerse test | material accuracy: 96.3% -> 93.9% | Hurt |
| train a generative model over softest-to-stiffest spectrum instead of deterministic midpoint regression | PIXIE · PixieMultiVerse test | log E MSE: 0.0205 -> 0.0091 | Improved |
| generative spectrum training instead of deterministic regression | PIXIE · PixieMultiVerse test | material accuracy: 97.3% -> 93.9% | Hurt |
| single unified model instead of specialist for MPM | PIXIE · MPM, alpha=0.0 | PSNR: 23.16 -> 29.25 | Improved |
| single unified model instead of specialist for MPM | PIXIE · MPM, alpha=0.5 | SSIM: 0.9225 -> 0.9198 | Hurt |
| single unified model instead of Vid2Sim (full) for LBS | Vid2Sim (full) · LBS, alpha=0.0 | PSNR: 28.30 -> 33.83 | Improved |
| single unified model instead of Vid2Sim (full) for LBS | Vid2Sim (full) · LBS, alpha=0.5 | PSNR: 36.94 -> 36.81 | No clear effect |
| single unified model instead of tuned Spring-Gaus | Spring-Gaus (tuned) · Spring-Mass, alpha=0.0 | PSNR: 30.53 -> 33.88 | Improved |
| feed-forward prediction instead of test-time optimization | Spring-Gaus · Spring-Mass | runtime (s): 4375.30 -> 21.64 | Improved |
Closest papers by meaning
- UniPhy: Learning a Unified Constitutive Model for Inverse Physics Simulation (CVPR 2025 · similarity 0.90)
- OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025 · similarity 0.88)
- PhysX-3D: Physical-Grounded 3D Asset Generation (NeurIPS 2025 · similarity 0.87)
- Latent Intuitive Physics: Learning to Transfer Hidden Physics from A 3D Video (ICLR 2024 · similarity 0.87)
- DEL: Discrete Element Learner for Learning 3D Particle Dynamics with Neural Rendering (NeurIPS 2024 · similarity 0.86)
- PhyRecon: Physically Plausible Neural Scene Reconstruction (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
