Evaluating Reliability Gaps in Large Language Model Safety via Repeated Prompt Sampling

A connected record in the developer community graph.