Adversarial Representation Engineering: A General Model Editing Framework for Large Language Models

TL;DR

This paper introduces a novel framework called Adversarial Representation Engineering (ARE) that enables flexible, interpretable editing of large language models by leveraging internal representations and a robust sensor, improving model adaptability without performance loss.

Contribution

The paper proposes a unified, interpretable model editing framework using adversarial representation engineering and a sensor oracle, addressing challenges in practical large language model editing.

Findings

Effective model editing demonstrated across multiple tasks

Maintains baseline performance after editing

Provides a robust and reliable sensor for model manipulation

Abstract

Since the rapid development of Large Language Models (LLMs) has achieved remarkable success, understanding and rectifying their internal complex mechanisms has become an urgent issue. Recent research has attempted to interpret their behaviors through the lens of inner representation. However, developing practical and efficient methods for applying these representations for general and flexible model editing remains challenging. In this work, we explore how to leverage insights from representation engineering to guide the editing of LLMs by deploying a representation sensor as an editing oracle. We first identify the importance of a robust and reliable sensor during editing, then propose an Adversarial Representation Engineering (ARE) framework to provide a unified and interpretable approach for conceptual model editing without compromising baseline performance. Experiments on multiple tasks demonstrate the effectiveness of ARE in various model editing scenarios. Our code and data are available at https://github.com/Zhang-Yihao/Adversarial-Representation-Engineering.