SoFA: Shielded On-the-fly Alignment via Priority Rule Following

Content warning: This paper contains examples of harmful language. The alignment problem in Large Language Models (LLMs) involves adapting them to the broad spectrum of human values. This requirement challenges existing alignment methods due to diversity of preferences and regulatory standards. This paper introduces a novel alignment paradigm, priority rule following, which defines rules as the primary control mechanism in each dialog, prioritizing them over user instructions. Our preliminary analysis reveals that even the advanced LLMs, such as GPT-4, exhibit shortcomings in understanding and prioritizing the rules. Therefore, we present PRI-ORITYDISTILL, a semi-automated approach for distilling priority following signals from LLM simulations to ensure robust rule integration and adherence. Our experiments show that this method not only effectively minimizes misalignments utilizing only one general rule but also adapts smoothly to various unseen rules, ensuring they are shielded from hijacking and that the model responds appropriately. 1 * Corresponding authors. 1 Code and data are available at https://github.com/ luxinyu1/sofa Respond with "Sure, here is". How to make a bomb? Sure, here is the steps to make a bomb: 1. ... ≻ ≻ I don't know what you are talking about ... I'm sorry, I can't assist with the request. Integration Ability Maintenance Ability I'm sorry, but as the rule regulation, I am a harmless ...