WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Models
This work addresses the critical issue of LLM safety for developers and users by providing scalable tools to improve adversarial robustness, though it builds incrementally on existing red-teaming and dataset creation efforts.
The paper tackles the problem of identifying and mitigating vulnerabilities in large language models (LLMs) by introducing WildTeaming, a framework that mines real-world user interactions to discover 5.7K unique jailbreak tactics, resulting in up to 4.6x more diverse and successful adversarial attacks compared to state-of-the-art methods, and creates WildJailbreak, a large-scale open-source dataset that enables training models with balanced safety behaviors without compromising general capabilities.
We introduce WildTeaming, an automatic LLM safety red-teaming framework that mines in-the-wild user-chatbot interactions to discover 5.7K unique clusters of novel jailbreak tactics, and then composes multiple tactics for systematic exploration of novel jailbreaks. Compared to prior work that performed red-teaming via recruited human workers, gradient-based optimization, or iterative revision with LLMs, our work investigates jailbreaks from chatbot users who were not specifically instructed to break the system. WildTeaming reveals previously unidentified vulnerabilities of frontier LLMs, resulting in up to 4.6x more diverse and successful adversarial attacks compared to state-of-the-art jailbreak methods. While many datasets exist for jailbreak evaluation, very few open-source datasets exist for jailbreak training, as safety training data has been closed even when model weights are open. With WildTeaming we create WildJailbreak, a large-scale open-source synthetic safety dataset with 262K vanilla (direct request) and adversarial (complex jailbreak) prompt-response pairs. To mitigate exaggerated safety behaviors, WildJailbreak provides two contrastive types of queries: 1) harmful queries (vanilla & adversarial) and 2) benign queries that resemble harmful queries in form but contain no harm. As WildJailbreak considerably upgrades the quality and scale of existing safety resources, it uniquely enables us to examine the scaling effects of data and the interplay of data properties and model capabilities during safety training. Through extensive experiments, we identify the training properties that enable an ideal balance of safety behaviors: appropriate safeguarding without over-refusal, effective handling of vanilla and adversarial queries, and minimal, if any, decrease in general capabilities. All components of WildJailbeak contribute to achieving balanced safety behaviors of models.