Dylan Everingham

Revisiting Cauchy-Schwarz Norm Bounds for SDC Detection in LLM Inference

Large Language Models (LLMs) execute billions of floating-point operations per output token across many hardware accelerators. At this scale, hardware-induced faults can bypass system-level detection mechanisms and cause Silent Data Corruption (SDC). Algorithm-Based Fault Tolerance (ABFT) is the state-of-the-art techni...