1 paper · 1 filter
Anton Altenbernd, Dylan Everingham, Philipp Wiesner +1
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 sys…