Abstract Recent advances in AI are pushing the limits of traditional hardware, making physical neural networks (PNNs) a promising alternative. However, training PNNs remains challenging: in silico training suffers from model-reality mismatch, while in situ training produces device-specific models that do not transfer across fabrication variations. Both approaches are further compromised by post-deployment perturbations, such as thermal drift or misalignment, which invalidate trained models and require retraining. We address these challenges through sharpness-aware training (SAT), inspired by sharpness-aware minimization, which links loss landscape geometry to generalization. We establish a connection between loss landscape sharpness and robustness in physical systems and leverage it to improve PNN training. SAT is compatible with both in silico and in situ settings: it mitigates model-reality gaps, enables cross-device transfer, and provides strong resilience to post-deployment perturbations without retraining. We demonstrate SAT across three PNN platforms and multiple tasks, including classification, compression, reconstruction, and generation, showing its broad applicability.
Xu et al. (Mon,) studied this question.