Research map / Probabilistic and scientific modeling / Dynamical systems
Interpretable Intuitive Physics Model
Tian Ye, Xiaolong Wang, James Davidson, Abhinav Gupta
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 disentanglement constraints on bottleneck (interpretable model) vs end-to-end baseline (ablation) | Intuitive physics model · Unseen shape combinations | pixel prediction error: 95.6 -> 87.3 | Improved |
| add disentanglement constraints on bottleneck vs baseline (ablation) | Intuitive physics model · Interpolation shape 5 | pixel prediction error: 173.80 -> 138.04 | Improved |
| add disentanglement constraints on bottleneck vs baseline (ablation) | Intuitive physics model · Interpolation parameter 3 | pixel prediction error: 299.88 -> 154.09 | Improved |
| add disentanglement constraints on bottleneck vs baseline (ablation) | Intuitive physics model · Ratio parameter ratio 3 | pixel prediction error: 490.92 -> 157.10 | Improved |
| add disentanglement constraints on bottleneck vs baseline (ablation) | Intuitive physics model · Ratio shape ratio 2 | pixel prediction error: 345.60 -> 110.79 | Improved |
| predict from ground-truth physics parameters and optical flows (Flow + Physics) instead of learned interpretable latent (ablation) | Prediction model · Interpolation parameter 3 | pixel prediction error: 154.09 -> 671.51 | Hurt |
| predict via supervised optical flow of frames 4-5 instead of pixels of 5th frame (ablation) | Prediction model · Unseen shape combinations | pixel error: 87.3 -> 118.8 | Hurt |
| use linear model on optical flows instead of the learned model (ablation) | Prediction model · Unseen shape combinations | pixel error: 87.3 -> 292.5 | Hurt |
| use staggered training (curriculum over properties) (ablation) | Intuitive physics model · Training | not reported | Improved |
Closest papers by meaning
- Physion++: Evaluating Physical Scene Understanding that Requires Online Inference of Different Physical Properties (NeurIPS 2023 · similarity 0.87)
- Modeling Expectation Violation in Intuitive Physics with Coarse Probabilistic Object Representations (NeurIPS 2019 · similarity 0.86)
- Latent Intuitive Physics: Learning to Transfer Hidden Physics from A 3D Video (ICLR 2024 · similarity 0.85)
- Scaling Physical Reasoning with the PHYSICS Dataset (NeurIPS 2025 · similarity 0.85)
- Physical Property Understanding from Language-Embedded Feature Fields (CVPR 2024 · similarity 0.83)
- Beyond Static Vision: Scene Dynamic Field Unlocks Intuitive Physics Understanding in Multi-modal Large Language Models (ICLR 2026 · similarity 0.83)
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.
