Characteristics of Harmful Text: Towards Rigorous Benchmarking of Language Models

Open in new window