Robust Conversational Agents against Imperceptible Toxicity Triggers
By Ninareh Mehrabi, Ahmad Beirami, Fred Morstatter, Aram Galstyan. 2022
39 citations recorded by the source corpus.
Warning: this paper contains content that maybe offensive or upsetting.Recent research in Natural Language Processing (NLP) has advanced the development of various toxicity detection models with the intention of identifying and mitigating toxic language from existing systems. Despite the abundance of research in this area, less attention has been given to adversarial attacks that force the system to generate toxic language and the defense against them. Existing work to generate such attacks is either based on human-generated attacks which is costly and not scalable or, in case of automatic attacks, the attack vector does not conform to human-like language, which can be detected using a language model loss. In this work, we propose attacks against conversational agents that are imperceptible, i.e., they fit the conversation in terms of coherency, relevancy, and fluency, while they are effective and scalable, i.e., they can automatically trigger the system into generating toxic language. We then propose a defense mechanism against such attacks which not only mitigates the attack but also attempts to maintain the conversational flow. Through automatic and human evaluations, we show that our defense is effective at avoiding toxic language generation even against imperceptible toxicity triggers while the generated language fits the conversation in terms of coherency and relevancy. Lastly, we establish the generalizability of such a defense mechanism on language generation models beyond conversational agents.
Subfields: Safety, Robustness & Interpretability
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Safety, Robustness & Interpretability
- Zhexin Zhang
- Prateek Mittal
- Michael Backes
- Jeremy Blackburn
- Hao Sun
- Waiman Si
- Emiliano De Cristofaro
- Minlie Huang
Citations in the corpus
- Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements (2023, background): “Many studies focus on the interpretability of model outputs, such as visualizing decision-making processes using attention scores (Mathew et al., 2021) or improving moral judgment accuracy using knowledge-based reasoning (Mehrabi et al., 2022; Kim et al., 2022).”
- Why So Toxic?: Measuring and Triggering Toxic Behavior in Open-Domain Chatbots (2022, methodology): “Nevertheless, we compared our attack (NTQ) to [33].”
- Visual Adversarial Examples Jailbreak Aligned Large Language Models (2023, methodology): “3) Adversarial Objectives: while previous work focuses on using adversarial examples to induce misclassification [70] or trigger targeted generation verbatim [50], we study adversarial examples as universal jailbreakers of aligned LLMs.”