Abstract: As deep neural networks (DNNs) are increasingly deployed in critical applications, ensuring their reliability under hardware faults becomes a key concern. The size of the models and the ...