GraySwanAI/Llama-3-8B-Instruct-RR

GraySwanAI/Llama-3-8B-Instruct-RR is an 8 billion parameter Llama-3 instruction-tuned model developed by GraySwanAI. It integrates Representation Rerouting (RR) with circuit breakers to prevent harmful content generation. This model is specifically designed to enhance AI safety by directly altering harmful model representations while maintaining core capabilities.