Phys-JEPA: Physics-Informed Latent World Models for Multivariate Time-Series Forecasting
For researchers in physics-informed machine learning and time-series forecasting, this work proposes a new paradigm for integrating physical constraints into latent representations, but the improvements are incremental and not yet SOTA.
Phys-JEPA introduces a physics-informed latent world model for multivariate time-series forecasting, where physical consistency is enforced on latent states and transitions rather than only on decoded outputs. On Jena Climate, it reduces aggregate MSE from 0.12482 to 0.12273 at H=24; on Traffic, it improves H=192 MSE from 0.800784 to 0.773873; on Electricity, results vary by horizon.
Multivariate forecasting in physical systems requires models that predict coupled temporal variables while preserving meaningful state evolution. Deep forecasters can fit temporal correlations, and physics-informed models can regularize predictions with scientific constraints, but these directions are often connected only at the decoded-output level. As a result, the hidden predictive state that generates future trajectories may remain statistically useful but physically unstructured. We introduce Phys-JEPA, a physics-informed joint-embedding predictive architecture for multivariate time-series forecasting. Phys-JEPA learns a latent world model in which predictive states are decomposed into physical and residual components, and physical consistency is imposed directly on latent states and latent transitions rather than only on decoded forecasts. This formulation uses known physical variables to organize the representation space while retaining residual capacity for unresolved dynamics. On Jena Climate 2009--2016, Phys-JEPA reduces aggregate MSE from 0.12482 to 0.12273 and temperature MSE from 0.01892 to 0.01831 at H=24. On Traffic, full Phys-JEPA improves aggregate MSE over the supervised baseline across all tested horizons, reducing H=192 MSE from 0.800784 to 0.773873. On Electricity, the best variant depends on horizon: static latent consistency is strongest at H=24 and H=48, while full Phys-JEPA gives the best aggregate and target-variable MSE at H=192. These initial results suggest that moving physics-informed learning from output space to latent predictive state space is a promising direction for interpretable temporal world models.