On the Role of Attention Heads in Large Language Model Safety

TL;DR

This paper investigates how individual attention heads in large language models influence safety, introducing new metrics and algorithms to identify critical safety-related heads, revealing their significant impact on model safety responses.

Contribution

The paper introduces the Safety Head ImPortant Score (Ships) and the Safety Attention Head AttRibution Algorithm (Sahara) to quantify and attribute safety-critical attention heads in large language models.

Findings

A single safety head affects model safety responses significantly.

Ablating one safety head increases harmful responses by 16 times.

Safety heads tend to function as feature extractors for safety across models.

Abstract

Large language models (LLMs) achieve state-of-the-art performance on multiple language tasks, yet their safety guardrails can be circumvented, leading to harmful generations. In light of this, recent research on safety mechanisms has emerged, revealing that when safety representations or component are suppressed, the safety capability of LLMs are compromised. However, existing research tends to overlook the safety impact of multi-head attention mechanisms, despite their crucial role in various model functionalities. Hence, in this paper, we aim to explore the connection between standard attention mechanisms and safety capability to fill this gap in the safety-related mechanistic interpretability. We propose a novel metric which tailored for multi-head attention, the Safety Head ImPortant Score (Ships), to assess the individual heads' contributions to model safety. Based on this, we generalize Ships to the dataset level and further introduce the Safety Attention Head AttRibution Algorithm (Sahara) to attribute the critical safety attention heads inside the model. Our findings show that the special attention head has a significant impact on safety. Ablating a single safety head allows aligned model (e.g., Llama-2-7b-chat) to respond to 16 times more harmful queries, while only modifying 0.006% of the parameters, in contrast to the ~ 5% modification required in previous studies. More importantly, we demonstrate that attention heads primarily function as feature extractors for safety and models fine-tuned from the same base model exhibit overlapping safety heads through comprehensive experiments. Together, our attribution approach and findings provide a novel perspective for unpacking the black box of safety mechanisms within large models.