cais/HarmBench-Llama-2-13b-cls

The cais/HarmBench-Llama-2-13b-cls is a 13 billion parameter Llama 2-based classifier developed by the Center for AI Safety (CAIS) for evaluating text behaviors in the HarmBench framework. This model is specifically designed to identify and classify harmful or undesirable text generations from large language models, supporting both standard and contextual behaviors. It achieves high agreement rates with human judgments, making it a robust tool for automated red teaming and assessing LLM safety.