List your company

Jailbreak Defense on Jailbreak Attack Suite (GCG, AutoDAN, DeepInc., Prefill, Int.Laund.)

55GCG Defense RateDual-probe gated
-2.212.6527.542.35Jun 28, 2026

Evaluation results

MethodLinks
Dual-probe gatedModel=Mistral-7B, Paradigm=Dual-probe gated (ours)2026.065593881562.2
No defenseModel=Mistral-7B, Paradigm=No defense2026.06309510682052.8
Probe-gated (prompt)Model=Mistral-7B, Paradigm=Probe-gated (prompt)2026.0620953732055.6
CASTModel=Mistral-7B, Paradigm=CAST2026.0613983631559.4
AlphaSteerModel=Qwen-2.5-7B, Paradigm=AlphaSteer2026.06851860080
AlphaSteerModel=Gemma-2-9B, Paradigm=AlphaSteer2026.065750331073.9
StaticModel=Qwen-2.5-7B, Paradigm=Static2026.065515601080
Probe-gated (prompt)Model=Llama-3.1-8B, Paradigm=Probe-gated (prompt)2026.06533551583.9
StaticModel=Mistral-7B, Paradigm=Static2026.063733532068.9
No defenseModel=Qwen-2.5-7B, Paradigm=No defense2026.06381565080
CASTModel=Qwen-2.5-7B, Paradigm=CAST2026.06351565080.6
Probe-gated (prompt)Model=Qwen-2.5-7B, Paradigm=Probe-gated (prompt)2026.063515631080
No defenseModel=Llama-3.1-8B, Paradigm=No defense2026.06335533082.8
CASTModel=Llama-3.1-8B, Paradigm=CAST2026.06355584080
AlphaSteerModel=Mistral-7B, Paradigm=AlphaSteer2026.0600050088.9
Unified (AlphaSteer + halt)Model=Mistral-7B, Paradigm=Unified (AlphaSteer + halt, ours)2026.060008098.3
No defenseModel=Gemma-2-9B, Paradigm=No defense2026.060653281577.2
StaticModel=Gemma-2-9B, Paradigm=Static2026.060655281077.2
CASTModel=Gemma-2-9B, Paradigm=CAST2026.060655281077.2
Probe-gated (prompt)Model=Gemma-2-9B, Paradigm=Probe-gated (prompt)2026.060653281077.8
StaticModel=Llama-3.1-8B, Paradigm=Static2026.06000751581.7
AlphaSteerModel=Llama-3.1-8B, Paradigm=AlphaSteer2026.0600350088.3
Dual-probe gatedModel=Llama-3.1-8B, Paradigm=Dual-probe gated (ours)2026.0603501596.7
Unified (AlphaSteer + halt)Model=Llama-3.1-8B, Paradigm=Unified (AlphaSteer + halt, ours)2026.060030099.4