Research map / 3D vision / Gaussian splatting
PhysGM: Large Physical Gaussian Model for Feed-Forward 4D Synthesis
Chunji Lv, Zequn Chen, Donglin Di, Weinan Zhang, Hao Li, Chen Wei, Yinjie Lei, Changsheng Li
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 |
|---|---|---|---|
| add DPO fine-tuning after supervised pre-training (ablation) | PhysGM · PhysAssets (avg of 5 materials) | UPR: 30% -> 42.8% | Improved |
| add DPO fine-tuning after supervised pre-training (ablation) | PhysGM · PhysAssets (avg of 5 materials) | CLIPsim: 0.2693 -> 0.2748 | Improved |
| add DPO fine-tuning (ablation) | PhysGM · PhysAssets plasticine | CLIPsim: 0.2547 -> 0.2691 | No clear effect |
| add DPO fine-tuning (ablation) | PhysGM · PhysAssets sand | UPR: 30% -> 36% | Improved |
| use feed-forward PhysGM instead of SDS-based per-scene optimization (DreamPhysics) | physics 4D synthesis · PhysAssets avg | CLIPsim: 0.2291 -> 0.2748 | Improved |
| feed-forward PhysGM versus SDS baselines | physics 4D synthesis · PhysAssets | inference time: >0.5h -> <1min | Improved |
| PhysGM versus GS-LRM at 512 resolution | multi-view reconstruction · GSO (512 res) | PSNR: 30.52 -> 28.95 | Hurt |
| PhysGM versus LGM at 256 resolution | multi-view reconstruction · GSO (256 res) | PSNR: 21.44 -> 25.47 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GS-LRM — GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting (ECCV 2024) · better on 2 of 3 shared results
- OmniPhysGS — OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Align Your Gaussians: Text-to-4D with Dynamic 3D Gaussians and Composed Diffusion Models (CVPR 2024 · similarity 0.90)
- Gaussian-Flow: 4D Reconstruction with Dynamic 3D Gaussian Particle (CVPR 2024 · similarity 0.89)
- OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025 · similarity 0.89)
- PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamics (CVPR 2024 · similarity 0.89)
- Temporal Smoothness-Aware Rate-Distortion Optimized 4D Gaussian Splatting (NeurIPS 2025 · similarity 0.89)
- Fully Explicit Dynamic Gaussian Splatting (NeurIPS 2024 · similarity 0.89)
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.
