Adversarial Arena: Crowdsourcing Data Generation through Interactive Competition

Post-training Large Language Models requires diverse, high-quality data which is rare and costly to obtain, especially in low resource domains and for multi-turn conversations. Common solutions are crowdsourcing or synthetic generation, but both often yield low-quality or low-diversity data. We introduce Adversarial Arena for building high quality conversational datasets by framing data generation as an adversarial task: attackers create prompts, and defenders generate responses. This interactive competition between multiple teams naturally produces diverse and complex data. We validated this approach by conducting a competition with 10 academic teams from top US and European universities, each building attacker or defender bots. The competition, focused on safety alignment of LLMs in cybersecurity, generated 19,683 multi-turn conversations. Fine-tuning an open-source model on this dataset produced an 18.47% improvement in secure code generation on CyberSecEval-Instruct and 29.42% improvement on CyberSecEval-MITRE.

Key Contributions

• Adversarial Arena framework for generating high-quality multi-turn conversational datasets through competitive red-teaming
• Cybersecurity safety alignment dataset of 19,683 multi-turn conversations from 10 academic teams
• 18.47% improvement on CyberSecEval-Instruct and 29.42% improvement on CyberSecEval-MITRE benchmarks

🛡️ Threat Analysis

Details

Similar Papers