The paper investigates how two independent inductive biases—one from the circuit’s invariance under conductance rescaling and one from the learning rule’s conservation of a mass quantity—affect what a physical learning system remembers. By separating these effects, the authors show that when every element is trainable, the initialization scale has negligible influence on the learned function, whereas a single untrainable element can cause the function to shift significantly with initialization. They further demonstrate that the conservation law does not protect memory but instead influences solution quality, with adjoint coupled learning (AL) generally performing worse than equilibrium propagation (EP) and coupled learning (CL) in small circuits.
whyItMatters":"The study clarifies that only the circuit’s structural bias, not the rule’s conservation property, determines memory retention in physical learning systems."
By Bijaya Dangol
arXiv:2606. 15444v1 Announce Type: cross Abstract: In this paper we show that the physical learning methods known as coupled learning (CL) and equilibrium propagation (EP) conserve a mass-like quantity in the trainable parameters in the continuous-time, small-nudging limit.
By Joshua A. McGinnis, Adam G. Kline, Yoichiro Mori
arXiv:2606. 15443v1 Announce Type: cross Abstract: Physical learning methods train physical networks to perform computational tasks using only local update rules, exploiting the physics of the system to handle the global transfer of information.
By Joshua A. McGinnis, Xinbo Li, Yoichiro Mori
arXiv:2608.30778v1 Announce Type: new
Abstract: Physical learning lets a trainable material or network use its own physical response to carry error signals, reducing the need for a separately program...
By Ruiwu Niu, Xiaowen Bi, Micha\"el Antonie van Wyk
The paper demonstrates that learned simulators can fail in two distinct ways when conditions change: long‑horizon drift due to accumulated errors and incorrect responses to interventions on physical parameters. By adding a symplectic integrator to preserve conservative dynamics, rollouts remain stable for up to 100× the training horizon, while encoding physical coupling via explicit linear factorization allows the model to generalize to unseen signs of that coupling. The study shows that stability and counterfactual generalization arise from separate structural choices, enabling designers to impose each property independently.
By Yufeng Wang, Parivesh Priye, Lu Wei, Haibin Ling
arXiv:2602. 03846v2 Announce Type: replace-cross Abstract: We develop a continual learning method for pretrained models that \emph{requires no access to old-task data}, addressing a practical barrier in foundation model adaptation where pretraining distributions are often unavailable.
By Romain Cosentino
arXiv:2606. 09929v1 Announce Type: cross Abstract: Physical reservoir computing harnesses nonlinear mechanical dynamics but, by convention, freezes the substrate and trains only a linear readout, presuming the substrate is not usefully trainable.
By Caleb Munigety
arXiv:2609.38768v1 Announce Type: cross
Abstract: Prior work has shown that neural networks exhibit implicit biases toward low-complexity structure (e.g., spectral bias), memorization dynamics, and c...
By Fujie Gao, Zuyue Zhang, Gang Sun
The paper introduces a world model that learns to predict the evolution of physical systems while respecting key physical principles. By hard‑coding a general structure—generating dynamics from the gradient of a learned energy via a fixed reversible operator and imposing constraints on energy, dissipation, and interventions—the model achieves second‑law compatible dissipation, accurate responses to parameter changes, long‑term stability, and robustness to disturbances. Experiments on an electromagnetic cavity, a particle‑in‑cell grid, and shallow‑water fluid demonstrate that the model can recover accurate constitutive functions, distinguish conserving from dissipating regimes, and transfer learned physics to unseen conditions, outperforming unconstrained models.
By Yufeng Wang, Parivesh Priye, Lu Wei, Haibin Ling
Prior work has shown that neural networks exhibit implicit biases toward low-complexity structure (e.g., spectral bias), memorization dynamics, and compression-like effects during training, but a unif...
arXiv:2608.23889v1 Announce Type: cross
Abstract: Plasticity under changing environments is central to both evolutionary biology and continual learning. Motivated by recent work on genotype--phenotyp...
By Owen Skriloff
arXiv:2606. 29519v1 Announce Type: new Abstract: Long-range learning is hard for recurrent networks trained with stochastic gradient descent, because the influence of a past input fades with the lag $\ell$, and if it fades too fast the dependence cannot be learned from finite data.
By Lorenzo Livi