Patching LLM Like Software: A Lightweight Method for Improving Safety Policy in Large Language Models

Open in new window