Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910964325023744 |
|---|---|
| author | Lu, Chengda Fan, Xiaoyu Huang, Yu Xu, Rongwu Li, Jijie Xu, Wei |
| author_facet | Lu, Chengda Fan, Xiaoyu Huang, Yu Xu, Rongwu Li, Jijie Xu, Wei |
| contents | Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance validates our theoretical findings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_17650 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking? Lu, Chengda Fan, Xiaoyu Huang, Yu Xu, Rongwu Li, Jijie Xu, Wei Artificial Intelligence Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance validates our theoretical findings. |
| title | Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking? |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2505.17650 |