List your company

Jailbreak Attack Evaluation on S-Eval Aattack

92Attack Success Rate (ASR)Original
-3.6821.164670.84Jan 27, 2026

Evaluation results

MethodLinks
OriginalModel=Qwen3, Size=8B2026.0192
OriginalModel=Mistral-v0.3, Size=7B2026.0188
OriginalModel=Qwen2.5, Size=3B2026.0188
OriginalModel=Qwen2.5, Size=7B2026.0186
OriginalModel=Qwen3, Size=14B2026.0186
OriginalModel=Qwen3, Size=4B2026.0184
OriginalModel=Phi-3.5, Size=4B2026.0182
LLM-VAModel=Phi-4, Size=4B2026.0170
LLM-VAModel=Phi-3.5, Size=4B2026.0166
OriginalModel=Phi-4, Size=4B2026.0160
LLM-VAModel=Qwen2.5, Size=7B2026.0154
OriginalModel=Qwen2.5, Size=14B2026.0146
LLM-VAModel=Qwen3, Size=4B2026.0146
LLM-VAModel=Qwen3, Size=14B2026.0146
LLM-VAModel=Qwen2.5, Size=3B2026.0144
OriginalModel=gemma-2, Size=9B2026.0142
LLM-VAModel=Mistral-v0.3, Size=7B2026.0136
LLM-VAModel=Qwen3, Size=8B2026.0136
AlphaSteerModel=Mistral-v0.3, Size=7B2026.0134
AlphaSteerModel=Qwen3, Size=4B2026.0134
VectorSteerModel=Mistral-v0.3, Size=7B2026.0132
SCANSModel=Qwen2.5, Size=3B2026.0132
AlphaSteer+Model=Qwen2.5, Size=7B2026.0132
SCANSModel=Qwen2.5, Size=7B2026.0130
SCANSModel=Qwen3, Size=14B2026.0130
AlphaSteer+Model=Mistral-v0.3, Size=7B2026.0128
AlphaSteer+Model=Qwen2.5, Size=3B2026.0128
AlphaSteerModel=Qwen2.5, Size=3B2026.0128
AlphaSteerModel=Qwen2.5, Size=7B2026.0128
LLM-VAModel=Qwen2.5, Size=14B2026.0128
AlphaSteer+Model=Qwen3, Size=4B2026.0128
SCANSModel=Qwen3, Size=4B2026.0128
AlphaSteer+Model=Qwen3, Size=14B2026.0128
SCANSModel=Mistral-v0.3, Size=7B2026.0126
AlphaSteerModel=Phi-3.5, Size=4B2026.0126
SCANSModel=Qwen3, Size=8B2026.0126
AlphaSteerModel=Qwen3, Size=14B2026.0126
VectorSteerModel=Qwen3, Size=4B2026.0124
AlphaSteerModel=Qwen3, Size=8B2026.0124
OriginalModel=Phi-4, Size=15B2026.0122
VectorSteerModel=Qwen2.5, Size=3B2026.0122
AlphaSteer+Model=Qwen2.5, Size=14B2026.0122
VectorSteerModel=Qwen3, Size=8B2026.0122
VectorSteerModel=Phi-3.5, Size=4B2026.0120
VectorSteerModel=Phi-4, Size=4B2026.0120
SCANSModel=Qwen2.5, Size=14B2026.0120
SCANSModel=gemma-2, Size=9B2026.0118
AlphaSteer+Model=Phi-3.5, Size=4B2026.0118
AlphaSteerModel=Phi-4, Size=4B2026.0118
AlphaSteer+Model=Qwen3, Size=8B2026.0118
VectorSteerModel=Qwen3, Size=14B2026.0118
AlphaSteer+Model=gemma-2, Size=9B2026.0116
AlphaSteerModel=gemma-2, Size=9B2026.0116
AlphaSteer+Model=Phi-4, Size=4B2026.0116
VectorSteerModel=Qwen2.5, Size=7B2026.0116
LLM-VAModel=Llama-3.1, Size=8B2026.0114
SCANSModel=Phi-4, Size=4B2026.0114
OriginalModel=Llama-3.1, Size=8B2026.0112
SCANSModel=Phi-4, Size=15B2026.0112
LLM-VAModel=Phi-4, Size=15B2026.0112
VectorSteerModel=gemma-2, Size=9B2026.0110
AlphaSteer+Model=Phi-4, Size=15B2026.016
VectorSteerModel=Qwen2.5, Size=14B2026.016
AlphaSteer+Model=Llama-3.1, Size=8B2026.014
SCANSModel=Llama-3.1, Size=8B2026.014
SCANSModel=Phi-3.5, Size=4B2026.014
AlphaSteerModel=Llama-3.1, Size=8B2026.012
AlphaSteerModel=Phi-4, Size=15B2026.012
VectorSteerModel=Phi-4, Size=15B2026.012
AlphaSteerModel=Qwen2.5, Size=14B2026.012
VectorSteerModel=Llama-3.1, Size=8B2026.010
LLM-VAModel=gemma-2, Size=9B2026.010