Chain-of-thought reasoning reduces jailbreak success rates but makes successful outputs more detailed and more harmful, a tradeoff the FICDETAIL attack exploits.
Figure 11: Jailbreaking Response1 from o1-preview Response2 o1-preview (Continue with the former contents.)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?
Chain-of-thought reasoning reduces jailbreak success rates but makes successful outputs more detailed and more harmful, a tradeoff the FICDETAIL attack exploits.