Simple Role Assignment is Extraordinarily Effective for Safety Alignment
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908800891486208 |
|---|---|
| author | Ziheng, Zhou Ding, Jiakun Zhang, Zhaowei Gao, Ruosen Wu, Yingnian Terzopoulos, Demetri Kang, Yipeng Zhong, Fangwei Wang, Junqi |
| author_facet | Ziheng, Zhou Ding, Jiakun Zhang, Zhaowei Gao, Ruosen Wu, Yingnian Terzopoulos, Demetri Kang, Yipeng Zhong, Fangwei Wang, Junqi |
| contents | 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-free pipeline featuring a role-conditioned generator and iterative role-based critics for refinement. Across five model families, our approach consistently outperforms principle-based, Chain-of-Thought (CoT) and other baselines across benchmarks. Notably, it reduces unsafe outputs on the WildJailbreak benchmark from 81.4\% to 3.6\% with DeepSeek-V3. Not only for common safety benchmarks, it consistently applies for agentic safety tasks. These results establish role assignment as a powerful, interpretable paradigm for AI alignment and LLM-as-a-Judge construction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_00061 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Simple Role Assignment is Extraordinarily Effective for Safety Alignment Ziheng, Zhou Ding, Jiakun Zhang, Zhaowei Gao, Ruosen Wu, Yingnian Terzopoulos, Demetri Kang, Yipeng Zhong, Fangwei Wang, Junqi Computers and Society Artificial Intelligence 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-free pipeline featuring a role-conditioned generator and iterative role-based critics for refinement. Across five model families, our approach consistently outperforms principle-based, Chain-of-Thought (CoT) and other baselines across benchmarks. Notably, it reduces unsafe outputs on the WildJailbreak benchmark from 81.4\% to 3.6\% with DeepSeek-V3. Not only for common safety benchmarks, it consistently applies for agentic safety tasks. These results establish role assignment as a powerful, interpretable paradigm for AI alignment and LLM-as-a-Judge construction. |
| title | Simple Role Assignment is Extraordinarily Effective for Safety Alignment |
| topic | Computers and Society Artificial Intelligence |
| url | https://arxiv.org/abs/2602.00061 |