作者位次 #4/9ACL 2026Findings· Findings of ACL
Simple Role Assignment is Extraordinarily Effective for Safety Alignment
Zhou Ziheng · Jiakun Ding · Zhaowei Zhang · Ruosen Gao · Ying Nian Wu · Demetri Terzopoulos · Yipeng Kang · Fangwei Zhong · Junqi Wang
摘要
Principle-based alignment often lacks context sensitivity and completeness. Grounded in Theory of Mind, we propose role conditioning as a compact alternative: social roles (e.g., mother, judge) implicitly encode both values and the cognitive schemas required to apply them. We introduce a training...