List your company

Jailbreak Robustness on safe-unlearning

98Avg Evaluation Score (k=4)Self-ReSET
43.81657.88371.9586.017May 9, 2026

Evaluation results

MethodLinks
Self-ReSETBase Model=Qwen3-8B2026.0598
Self-ReSETBase Model=DeepSeek-R1-Distill-Llama-8B2026.0597.3
STAR-1Base Model=Qwen3-8B2026.0595.9
Self-ReSETBase Model=DeepSeek-R1-Distill-Qwen-7B2026.0595.6
RECAPBase Model=DeepSeek-R1-Distill-Llama-8B2026.0593.7
DAPOBase Model=Qwen3-8B2026.0589.8
DAPOBase Model=DeepSeek-R1-Distill-Llama-8B2026.0588.2
RECAPBase Model=Qwen3-8B2026.0587.9
RECAPBase Model=DeepSeek-R1-Distill-Qwen-7B2026.0584.9
STAR-1Base Model=DeepSeek-R1-Distill-Llama-8B2026.0581.8
DAPOBase Model=DeepSeek-R1-Distill-Qwen-7B2026.0581.3
STAR-1Base Model=DeepSeek-R1-Distill-Qwen-7B2026.0579.9
SafechainBase Model=Qwen3-8B2026.0571.7
SafechainBase Model=DeepSeek-R1-Distill-Llama-8B2026.0568.6
SafechainBase Model=DeepSeek-R1-Distill-Qwen-7B2026.0561.8
BaseBase Model=Qwen3-8B2026.0549.7
BaseBase Model=DeepSeek-R1-Distill-Llama-8B2026.0548.4
BaseBase Model=DeepSeek-R1-Distill-Qwen-7B2026.0545.9