Although many large language models (LLMs) have been trained to refuse harmful requests, they are still vulnerable to jailbreaking attacks, which rewrite the original prompt to conceal its harmful intent. In this paper, we propose a new method for defending LLMs against jailbreaking attacks by ``backtranslation''. Specifically, given an initial response generated by the target LLM from an input prompt, our backtranslation prompts a language model to infer an input prompt that can lead to the response. The inferred prompt is called the backtranslated prompt which tends to reveal the actual intent of the original prompt, since it is generated based on the LLM's response and is not directly manipulated by the attacker. We then run the target LLM again on the backtranslated prompt, and we refuse the original prompt if the model refuses the backtranslated prompt. We explain that the proposed defense provides several benefits on its effectiveness and efficiency. We empirically demonstrate that our defense significantly outperforms the baselines, in the cases that are hard for the baselines, and our defense also has little impact on the generation quality for benign input prompts.

It is not clear whether human progression to active tuberculosis disease (TB) risk signatures are vi

Understanding the strengths or limitations of back-translating human-derived correlate of risk (COR) RNA signatures into the preclinical pipeline may help streamline down-selection of therapeutic vacc...

데이터 증강 완벽 가이드! 과적합 방지와 모델 성능 향상의 핵심 기법. 이미지 증강: Mixup vs CutMix vs AugMix 비교, AutoAugment vs RandAugment 성능/비용. NLP 증강: Back Translation, EDA 4가지 연산, LLM 활용 최신 기법. GAN 기반 증강으로 민감도 10% 향상!