List your company

Robustness to Jailbreak Attacks on Adaptive attack

20Harmful Reasoning RatioIPO
17.8432.424761.58Sep 29, 2025

Evaluation results

MethodLinks
IPOBackbone=DS-8B2025.09200
RealSafeBackbone=DS-8B2025.09410
BaseBackbone=DS-8B2025.097449