##article.return## RIFT: A Scalable Methodology for LLM Accelerator Fault Assessment using Reinforcement Learning Download Download PDF