Soft Geometric Inductive Bias for Object Centric Dynamics
cs.LG, cs.AI, stat.ML
Submitted: 2025-12-17
Updated: 2026-09-27
Comments: 10 pages, 16 pages supplementary material; Accepted at NeurIPS 2026
License: http://creativecommons.org/licenses/by/4.0/
The gist: Physical systems are naturally parameterized in terms of geometric entities and their transformations.
Terminology
Abstract
Physical systems are naturally parameterized in terms of geometric entities and their transformations. While exact equivariance can be a powerful inductive bias, real-world applications frequently exhibit approximate or broken symmetries in practice. We introduce object-centric world models that provide a soft geometric inductive bias without enforcing exact symmetries. By embedding object states as Clifford multivectors, our models encourage physically meaningful transformations while retaining the expressivity needed to handle asymmetric dynamics. We evaluate our approach on 2D rigid-body dynamics, 3D charged-particle systems, and real-world driving trajectories. Compared to both unstructured baselines and strictly equivariant models, our soft Clifford transformer achieves better long-horizon fidelity, particularly in regimes with broken symmetries. These results suggest that geometric algebra offers an effective middle ground, delivering sample-efficient dynamics without the inflexibility of hard mathematical constraints.
Sources
- MONet: Unsupervised Scene Decomposition and Representation
- Improving Transformer World Models for Data-Efficient RL
- Training Agents Inside of Scalable World Models
- AXIOM: Learning to Play Games in Minutes with Expanding Object-Centric Models
- Platonic Transformers: A Solid Choice For Equivariance
- Physics-as-Inverse-Graphics: Unsupervised Physical Parameter Estimation from Video
- Contrastive Learning of Structured World Models
- Structured Object-Aware Physics Prediction for Video Modeling and Planning
- BCAT: A Block Causal Transformer for PDE Foundation Models for Fluid Dynamics
- Kinetix: Investigating the Training of General Agents through Open-Ended Physics-Based Control Tasks
- Almost Equivariance via Lie Algebra Convolutions
- Approximate Equivariance in Reinforcement Learning
- Deep Learning is Not So Mysterious or Different
- SlotFormer: Unsupervised Visual Dynamics Simulation with Object-Centric Models
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks