Research map / Probabilistic and scientific modeling / Dynamical systems
VisionLaw: Inferring Interpretable Intrinsic Dynamics from Visual Observations via Bilevel Optimization
Jiajing Lin, Shu Jiang, Qingyuan Zeng, Zhenzhong Wang, Min Jiang
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 decoupled evolution (alternating elastic and plastic search before joint refinement) (ablation) | VisionLaw · Synthetic average | RGB loss: 10.73 -> 7.14 | Improved |
| add decoupled evolution (ablation) | VisionLaw · RubberPawn | RGB loss: 25.88 -> 15.25 | Improved |
| add decoupled evolution (ablation) | VisionLaw · ClayCat | RGB loss: 17.31 -> 8.13 | Improved |
| add decoupled evolution (ablation) | VisionLaw · Synthetic average | time (s): 11343 -> 11405 | No clear effect |
| add decoupled evolution (ablation) | VisionLaw · JellyDuck | time (s): 9333 -> 11249 | Hurt |
| use LLM-generated symbolic constitutive laws instead of neural constitutive law (NeuMA) | VisionLaw · Synthetic average | Chamfer distance: 2.86 -> 1.65 | Improved |
| use VisionLaw instead of NCLaw | VisionLaw · Synthetic average | Chamfer distance: 12.86 -> 1.65 | Improved |
| use VisionLaw instead of NeuMA | VisionLaw · SandFish | Chamfer distance: 1.07 -> 1.10 | Hurt |
| use VisionLaw instead of PAC-NeRF | VisionLaw · Synthetic average | Chamfer distance: 114.80 -> 1.65 | Improved |
| replace gradient-based lower-level optimization with evolutionary search (population 5 or 10) (ablation) | VisionLaw · Synthetic scenes | RGB loss | Mixed |
Closest papers by meaning
- Dynamics: Language-Based Representation for Inferring Rigid-Body Dynamics From Videos (CVPR 2026 · similarity 0.88)
- 3D-IntPhys: Towards More Generalized 3D-grounded Visual Intuitive Physics under Challenging Scenes (NeurIPS 2023 · similarity 0.87)
- Pixel2Phys: Distilling Governing Laws from Visual Dynamics (CVPR 2026 · similarity 0.87)
- PhysInOne: Visual Physics Learning and Reasoning in One Suite (CVPR 2026 · similarity 0.87)
- NeuMA: Neural Material Adaptor for Visual Grounding of Intrinsic Dynamics (NeurIPS 2024 · similarity 0.86)
- Mimicking the Physicist's Eye: A VLM-centric Approach for Physics Formula Discovery (NeurIPS 2026 · similarity 0.85)
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.
