Evaluating and Safeguarding the Adversarial Robustness of Retrieval-Based In-Context Learning

TL;DR

This paper investigates the adversarial robustness of retrieval-augmented in-context learning (ICL) with large language models, proposing a training-free defense method called DARD that improves robustness against attacks.

Contribution

It introduces DARD, a training-free adversarial defense method for retrieval-augmented ICL, and provides a comprehensive analysis of robustness against various adversarial attacks.

Findings

Retrieval-augmented ICL improves robustness against test sample attacks.

Overconfidence in demonstrations increases vulnerability to demonstration attacks.

DARD reduces attack success rate by 15% compared to baselines.

Abstract

With the emergence of large language models, such as LLaMA and OpenAI GPT-3, In-Context Learning (ICL) gained significant attention due to its effectiveness and efficiency. However, ICL is very sensitive to the choice, order, and verbaliser used to encode the demonstrations in the prompt. Retrieval-Augmented ICL methods try to address this problem by leveraging retrievers to extract semantically related examples as demonstrations. While this approach yields more accurate results, its robustness against various types of adversarial attacks, including perturbations on test samples, demonstrations, and retrieved data, remains under-explored. Our study reveals that retrieval-augmented models can enhance robustness against test sample attacks, outperforming vanilla ICL with a 4.87% reduction in Attack Success Rate (ASR); however, they exhibit overconfidence in the demonstrations, leading to a 2% increase in ASR for demonstration attacks. Adversarial training can help improve the robustness of ICL methods to adversarial attacks; however, such a training scheme can be too costly in the context of LLMs. As an alternative, we introduce an effective training-free adversarial defence method, DARD, which enriches the example pool with those attacked samples. We show that DARD yields improvements in performance and robustness, achieving a 15% reduction in ASR over the baselines. Code and data are released to encourage further research: https://github.com/simonucl/adv-retreival-icl