Evaluating and Safeguarding the Adversarial Robustness of Retrieval-Based In-Context Learning
This addresses security vulnerabilities in retrieval-augmented ICL for users of large language models, offering a practical defence method, though it is incremental as it builds on existing ICL and adversarial robustness research.
The paper tackles the adversarial robustness of retrieval-based in-context learning (ICL) in large language models, finding that retrieval-augmented models improve robustness against test sample attacks with a 4.87% reduction in Attack Success Rate (ASR) but are vulnerable to demonstration attacks with a 2% increase in ASR, and introduces a training-free defence method, DARD, that achieves a 15% reduction in ASR over baselines.
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