Learning to Poison Large Language Models During Instruction Tuning

The advent of Large Language Models (LLMs) has marked significant achievements in language processing and reasoning capabilities. Despite their advancements, LLMs face vulnerabilities to data poisoning attacks, where adversaries insert backdoor triggers into training data to manipulate outputs for malicious purposes. This work further identifies additional security risks in LLMs by designing a new data poisoning attack tailored to exploit the instruction tuning process. We propose a novel gradient-guided backdoor trigger learning approach to identify adversarial triggers efficiently, ensuring an evasion of detection by conventional defenses while maintaining content integrity. Through experimental validation across various LLMs and tasks, our strategy demonstrates a high success rate in compromising model outputs; poisoning only 1\% of 4,000 instruction tuning samples leads to a Performance Drop Rate (PDR) of around 80\%. Our work highlights the need for stronger defenses against data poisoning attack, offering insights into safeguarding LLMs against these more sophisticated attacks. The source code can be found on this GitHub repository: https://github.com/RookieZxy/GBTL/blob/main/README.md.

Ethical Considerations When Constructing Participatory Design Protocols for Social Robots

Participatory design has emerged as a popular approach to foreground ethical considerations in social robots by incorporating anticipated users and stakeholders as designers. Here we draw attention to the ethics of participatory design as a method, distinct from the ethical considerations of the social robot being co-designed. More specifically, we consider the ethical concerns posed by the act of stakeholder participation - the morals and values that should be explicitly considered when we, as researchers or practitioners, devise protocols for participatory design of social robots ("how" stakeholders participate). We use the case of robot-assisted sexual violence mitigation to exemplify ethical considerations of participatory design protocols such as risk of harm, exploitation, and reduction of stakeholder agency. To incorporate these and other ethical considerations in the creation of social robot participatory design protocols, we advocate letting stakeholders design their own form of participation by including them in the creation of participatory design sessions, structures, and processes.