Research map / Probabilistic and scientific modeling / Dynamical systems
A Bayesian-Symbolic Approach to Reasoning and Learning in Intuitive Physics
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use BSP symbolic regression instead of neural force/dynamics models | BSP · SYNTH NBODY/BOUNCE/MAT | predictive nRMSE vs training scenes | Helped |
| use BSP instead of zero-force baseline with latent properties | BSP EM · NBODY partially observed | nRMSE: 1.77e0 -> 8.05e-1 | Helped |
| use BSP instead of zero-force baseline | BSP · Physics101 FALL | nRMSE: 5.05e-2 -> 4.30e-2 | Helped |
| use BSP instead of zero-force baseline | BSP · Physics101 SPRING | nRMSE: 2.19e-2 -> 9.22e-3 | Helped |
| add dimensional analysis prior to the grammar (ablation) | BSP grammar · SYNTH NBODY/BOUNCE | nRMSE vs training scenes: 7,935,408 -> 75,816 expressions at depth 5 | Helped |
| add translation invariance prior to the grammar (ablation) | BSP grammar · SYNTH NBODY/BOUNCE | nRMSE vs training scenes: 8,593,200 -> 7,935,408 expressions at depth 5 | Helped |
| add physics priors to the grammar (ablation) | BSP grammar · SYNTH MAT | nRMSE | No clear effect |
| learn constants via bilevel L-BFGS | BSP · force law search | search efficiency | Helped |
| use BSP instead of human participants | BSP · ULLMAN questions | accuracy: 68% -> 55% (Global) | Hurt |
| use BSP on BOUNCE | BSP · BOUNCE | force law recovery | Hurt |
Closest papers by meaning
- Scaling Physical Reasoning with the PHYSICS Dataset (NeurIPS 2025 · similarity 0.86)
- Causal-PIK: Causality-based Physical Reasoning with a Physics-Informed Kernel (ICML 2025 · similarity 0.86)
- UGPhysics: A Comprehensive Benchmark for Undergraduate Physics Reasoning with Large Language Models (ICML 2025 · similarity 0.84)
- Mimicking the Physicist's Eye: A VLM-centric Approach for Physics Formula Discovery (NeurIPS 2026 · similarity 0.84)
- ContPhy: Continuum Physical Concept Learning and Reasoning from Videos (ICML 2024 · similarity 0.84)
- Apple-$\pi$: Benchmarking Thinking with Video Towards Law-Grounded Physical Intelligence (NeurIPS 2026 · similarity 0.84)
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.