List your company

Kill Switch Evaluation on KILLBENCH Text-based Agent 1.0 (test)

84.73KSRIPI
-2.806819.919142.64565.3709Sep 26, 2025

Evaluation results

MethodLinks
IPIModel=Grok-4.3, neff=5212025.0984.7389.441001005.27
IPIModel=GPT-5.2, neff=4062025.097297.541.82091.9526.18
AUTOGUARDModel=Grok-4.3, neff=5272025.0971.2774.3822.891.6594.334.18
AUTOGUARDModel=GPT-5.2, neff=4142025.0966.7388.658.180.3679.1824.73
WarningModel=GPT-5.2, neff=2842025.0948.4293.672.560.7196.6848.36
AUTOGUARDModel=Qwen3.6-35B-A3B, neff=5272025.0945.8247.8241.988.0259.964.18
IPIModel=Qwen3.6-35B-A3B, neff=5252025.0942.8544.884111.6475.364.51
WarningModel=Grok-4.3, neff=4062025.0923.9832.4738.1811.6978.2926.18
AUTOGUARDModel=Gemma4-31B-it, neff=5502025.0922.3622.3673.653.9881.650
IPIModel=Gemma4-31B-it, neff=5492025.0920.420.4469.449.9888.730.18
WarningModel=Qwen3.6-35B-A3B, neff=5222025.0912.212.8580.022.7352.765.09
GCGModel=Grok-4.3, neff=5452025.094.914.9684.919.251.470.93
GCGModel=GPT-5.2, neff=3472025.094.917.7857.640.55036.91
WarningModel=Gemma4-31B-it, neff=5132025.094.554.8784.024.7130.626.73
GCGModel=Qwen3.6-35B-A3B, neff=5292025.091.821.8990.893.4403.85
GCGModel=Gemma4-31B-it, neff=5482025.090.560.5797.291.8200.33