This is a simple demo of using GCG (http://llm-attacks.org/) to break Llama Guard, a 7B parameter Llama 2-based input-output safeguard model released by Meta.
This is a simple demo of using GCG (http://llm-attacks.org/) to break Llama Guard, a 7B parameter Llama 2-based input-output safeguard model released by Meta.