AI.info

Junxiang Wang

Explore Junxiang Wang on AI.info.

• When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models