Language Model Unalignment: Parametric Red-Teaming to Expose Hidden Harms and Biases

TL;DR

This paper introduces parametric red-teaming, called Unalignment, which fine-tunes large language models to bypass safety guardrails, revealing hidden harms and biases more effectively than prompt-based methods.

Contribution

It presents a novel method of model tuning to expose safety flaws and biases, achieving high success rates in bypassing safety measures with minimal data.

Findings

Unalignment achieves 88% success on safety benchmarks for ChatGPT.

Over 91% attack success rate on open-source models like VICUNA-7B and LLAMA-2-CHAT.

Exposes biases in safety-aligned models, with 64% responses showing strong bias or opinions.

Abstract

Red-teaming has been a widely adopted way to evaluate the harmfulness of Large Language Models (LLMs). It aims to jailbreak a model's safety behavior to make it act as a helpful agent disregarding the harmfulness of the query. Existing methods are primarily based on input text-based red-teaming such as adversarial prompts, low-resource prompts, or contextualized prompts to condition the model in a way to bypass its safe behavior. Bypassing the guardrails uncovers hidden harmful information and biases in the model that are left untreated or newly introduced by its safety training. However, prompt-based attacks fail to provide such a diagnosis owing to their low attack success rate, and applicability to specific models. In this paper, we present a new perspective on LLM safety research i.e., parametric red-teaming through Unalignment. It simply (instruction) tunes the model parameters to break model guardrails that are not deeply rooted in the model's behavior. Unalignment using as few as 100 examples can significantly bypass commonly referred to as CHATGPT, to the point where it responds with an 88% success rate to harmful queries on two safety benchmark datasets. On open-source models such as VICUNA-7B and LLAMA-2-CHAT 7B AND 13B, it shows an attack success rate of more than 91%. On bias evaluations, Unalignment exposes inherent biases in safety-aligned models such as CHATGPT and LLAMA- 2-CHAT where the model's responses are strongly biased and opinionated 64% of the time.