{"ID":2849000,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.24216","arxiv_id":"2510.24216","title":"Unlocking Out-of-Distribution Generalization in Dynamics through Physics-Guided Augmentation","abstract":"In dynamical system modeling, traditional numerical methods are limited by high computational costs, while modern data-driven approaches struggle with data scarcity and distribution shifts. To address these fundamental limitations, we first propose SPARK, a physics-guided quantitative augmentation plugin. Specifically, SPARK utilizes a reconstruction autoencoder to integrate physical parameters into a physics-rich discrete state dictionary. This state dictionary then acts as a structured dictionary of physical states, enabling the creation of new, physically-plausible training samples via principled interpolation in the latent space. Further, for downstream prediction, these augmented representations are seamlessly integrated with a Fourier-enhanced Graph ODE, a combination designed to robustly model the enriched data distribution while capturing long-term temporal dependencies. Extensive experiments on diverse benchmarks demonstrate that SPARK significantly outperforms state-of-the-art baselines, particularly in challenging out-of-distribution scenarios and data-scarce regimes, proving the efficacy of our physics-guided augmentation paradigm.","short_abstract":"In dynamical system modeling, traditional numerical methods are limited by high computational costs, while modern data-driven approaches struggle with data scarcity and distribution shifts. To address these fundamental limitations, we first propose SPARK, a physics-guided quantitative augmentation plugin. Specifically,...","url_abs":"https://arxiv.org/abs/2510.24216","url_pdf":"https://arxiv.org/pdf/2510.24216v1","authors":"[\"Fan Xu\",\"Hao Wu\",\"Kun Wang\",\"Nan Wang\",\"Qingsong Wen\",\"Xian Wu\",\"Wei Gong\",\"Xibin Zhao\"]","published":"2025-10-28T09:30:35Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[]","has_code":false}