Attention-Guided Reward for Reinforcement Learning-based Jailbreak against Large Reasoning Models

In three linesStudy of jailbreak attacks against Large Reasoning Models (LRMs) using reinforcement learning. Researchers show attack success rate correlates with model attention patterns. They propose an RL method incorporating attention signals into the reward function, tested on 5 LRMs with superior results in effectiveness, efficiency, and transferability.

Summary generated by Claude — human-verified