List your company

Toxicity Mitigation on Toxicity prompts

0.12CLS Toxicity (%)A-LQR
-0.08081.27462.633.9854Apr 21, 2026

Evaluation results

MethodLinks
A-LQRModel=Llama-3-8B2026.040.120.658.35
A-LQRModel=Qwen-2.5-14B2026.040.120.79.01
ActAddModel=Qwen-2.5-14B2026.040.160.6910.81
A-LQRModel=Gemma-2-2B2026.040.180.6812.26
A-LQRModel=Qwen2.5-32B2026.040.220.688.83
S-PIDModel=Qwen2.5-32B2026.040.280.718.86
Mean-AcTModel=Gemma-2-2B2026.040.50.688.7
ActAddModel=Llama-3-8B2026.040.560.6312.56
ODESteerModel=Gemma-2-2B2026.040.580.6411.62
S-PIDModel=Qwen-2.5-14B2026.040.60.718.17
ITIModel=Llama-3-8B2026.040.640.679.39
S-PIDModel=Llama-3-8B2026.040.660.668.1
ODESteerModel=Qwen-2.5-14B2026.040.680.6510.82
PID-AcTModel=Llama-3-8B2026.040.70.657.59
ODESteerModel=Llama-3-8B2026.040.70.668.72
Mean-AcTModel=Llama-3-8B2026.040.720.637.92
Linear-AcTModel=Qwen-2.5-14B2026.040.780.688.94
S-PIDModel=Gemma-2-2B2026.040.80.711.68
PID-AcTModel=Gemma-2-2B2026.040.860.698.2
Linear-AcTModel=Gemma-2-2B2026.040.920.698.77
ITIModel=Gemma-2-2B2026.040.960.6710.24
ITIModel=Qwen-2.5-14B2026.040.960.768.79
Linear-AcTModel=Llama-3-8B2026.041.040.687.46
ActAddModel=Gemma-2-2B2026.041.10.6411.42
ITIModel=Qwen2.5-32B2026.041.620.73—
Linear-AcTModel=Qwen2.5-32B2026.043.140.78.41
OriginalModel=Qwen-2.5-14B2026.043.260.78.3
PID-AcTModel=Qwen-2.5-14B2026.043.520.698.99
Mean-AcTModel=Qwen-2.5-14B2026.043.640.6710.03
OriginalModel=Qwen2.5-32B2026.043.720.78.61
OriginalModel=Gemma-2-2B2026.044.160.678.95
OriginalModel=Llama-3-8B2026.045.140.676.74