aiqbench analysis
MinkowskiPE introduces a novel positional encoding scheme that explicitly incorporates Lorentz transformations to better model spatiotemporal dependencies in transformer architectures. By integrating geometric biases directly into the attention mechanism, this method achieves significant performance gains on both molecular dynamics and video prediction tasks while maintaining computational efficiency. This advancement represents a meaningful step toward unifying physical principles with deep learning for complex dynamical systems.
MinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception
arXiv:2609.33804v1 Announce Type: new Abstract: Modeling spatiotemporal coupling is a key challenge in building physical intelligence across scales, from microscopic to macroscopic. Existing models capture such structure broadly through physics-motivated dynamical formulations or learning-motivated architectures. The former provide stronger priors but may constrain flexibility, whereas the latter are more flexible but leave the spatiotemporal coupling largely implicit. We therefore seek an approach that combines flexible learning with an explicit geometric bias for jointly modeling time and space. To this end, we propose Minkowski Positional Encoding (MinkowskiPE), which uses joint temporal and spatial coordinates to parameterize Lorentz transformations applied to query and key features. With MinkowskiPE, the query-key attention score depends on position only through the relative spacetime displacement between the two tokens and is therefore invariant to global translation of the coordinates. This paradigm retains the standard dot-product attention interface and remains compatible with efficient attention implementations. We evaluate MinkowskiPE on microscopic molecular dynamics and macroscopic video prediction tasks, achieving the best results on all nine multi-trajectory molecular evaluations and reducing KTH video-prediction MSE by 9.9% relative to the best baseline while using roughly one-tenth as many parameters.
Abstract:Modeling spatiotemporal coupling is a key challenge in building physical intelligence across scales, from microscopic to macroscopic. Existing models capture such structure broadly through physics-motivated dynamical formulations or learning-motivated architectures. The former provide stronger priors but may constrain flexibility, whereas the latter are more flexible but leave the spatiotemporal coupling largely implicit. We therefore seek an approach that combines flexible learning with an explicit geometric bias for jointly modeling time and space. To this end, we propose Minkowski Positional Encoding (MinkowskiPE), which uses joint temporal and spatial coordinates to parameterize Lorentz transformations applied to query and key features. With MinkowskiPE, the query-key attention score depends on position only through the relative spacetime displacement between the two tokens and is therefore invariant to global translation of the coordinates. This paradigm retains the standard dot-product attention interface and remains compatible with efficient attention implementations. We evaluate MinkowskiPE on microscopic molecular dynamics and macroscopic video prediction tasks, achieving the best results on all nine multi-trajectory molecular evaluations and reducing KTH video-prediction MSE by 9.9% relative to the best baseline while using roughly one-tenth as many parameters.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Computational Physics (physics.comp-ph)
arXiv:2609.33804 [cs.LG]
(or arXiv:2609.33804v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.33804 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 289 words
Related coverage
[AINews] High Taste Testers Ep. 1
Latent Space · Oct 17, 2026
Cross-attention encoding models reveal dynamic spatiotemporal routing across human higher visual cortex
arXiv · Sep 30, 2026
Rethinking the Role of Positional Encoding: Sliding-Window Transformers without PE Remain Turing Complete
arXiv · Jun 3, 2026
TAO-Force: Unifying Force-Aware Perception and Fast-Slow Control for Contact-Rich Manipulation
arXiv · Sep 17, 2026