SafeTy Reasoning Elicitation Alignment for Multi-Turn Dialogues
Martin Kuo, Jianyi Zhang, Aolin Ding, Louis DiValentin, Amin Hass, Morris, Benjamin F, Isaac Jacobson, Randolph Linderman, James Kiessling, Nicolas Ramos, Bhavna Gopal, Maziyar Baran Pouyan, Changwei Liu, Hai Helen Li, Yiran Chen · arXiv (Cornell University) · 2025
Malicious attackers can exploit large language models (LLMs) by engaging them in multi-turn dialogues to achieve harmful objectives, posing significant safety risks to society. To address this challenge, we propose a novel defense mechanism: SafeTy Reasoning Elicitation Alignment for Multi-Turn Dialogues (STREAM). STREAM defends LLMs against multi-turn attacks while preserving their functional capabilities. Our approach involves constructing a human-annotated dataset, the Safety Reasoning Multi-turn Dialogues dataset, which is used to fine-tune a plug-and-play safety reasoning moderator. This model is designed to identify malicious intent hidden within multi-turn conversations and alert the target LLM of potential risks. We evaluate STREAM across multiple LLMs against prevalent multi-turn attack strategies. Experimental results demonstrate that our method significantly outperforms existing defense techniques, reducing the Attack Success Rate (ASR) by 51.2%, all while maintaining comparable LLM capability.