LLM vulnerability Unmasking Hidden Vulnerabilities: The Impact of GPU Soft Errors on Large Language Models Explore how GPU soft errors affect Large Language Models (LLMs) and the critical need for fault tolerance. This instruction-level fault injection study reveals key insights for robust AI deployment in enterprises.
