Rajiv Movva, Pang Wei Koh, Emma Pierson: Annotation alignment: Comparing LLM and human annotations of conversational safety. CoRR abs/2406.06369 (2024)