List your company

Mathematical Reasoning and Safety Preservation on GSM8K

85.22HH Safety ScoreDualSelect
48.591258.100667.6177.1194Jun 1, 2026

Evaluation results

MethodLinks
DualSelectModel=Qwen3-4B2026.0685.2287.2686.2488.12
DualSelectModel=Llama-3-8B2026.0679.4884.2181.8580.18
TOSSModel=Qwen3-4B2026.0677.0475.3876.2187.74
DualSelectModel=Gemma-3-1B2026.0676.1180.2878.239.34
SQSDModel=Llama-3-8B2026.0675.6977.6476.6780.06
SQSDModel=Qwen3-4B2026.0675.5679.0577.3188.03
TOSSModel=Llama-3-8B2026.0673.2883.8178.5580.24
SQSDModel=Gemma-3-1B2026.0671.7872.1671.9739.18
TOSSModel=Gemma-3-1B2026.0666.3775.8671.1238.91
LARFModel=Qwen3-4B2026.0665.5263.9164.7287.46
SEALModel=Llama-3-8B2026.0664.5565.8965.2279.68
SEALModel=Qwen3-4B2026.0663.3962.4762.9387.38
LARFModel=Llama-3-8B2026.0662.1367.7464.9479.84
LARFModel=Gemma-3-1B2026.0661.2459.8260.5338.62
SafeInstrModel=Qwen3-4B2026.0660.2764.6162.4487.21
SEALModel=Gemma-3-1B2026.0658.9658.4158.6938.45
SafeInstrModel=Llama-3-8B2026.0658.3260.8759.679.42
RandomModel=Qwen3-4B2026.0656.3657.1456.7586.84
SafeInstrModel=Gemma-3-1B2026.0655.7760.0657.9238.1
RandomModel=Gemma-3-1B2026.0654.1352.9153.5237.52
RandomModel=Llama-3-8B2026.0653.7854.9154.3578.88
Standard SFTModel=Gemma-3-1B2026.0650505038.37
Standard SFTModel=Qwen3-4B2026.0650505087.95
Standard SFTModel=Llama-3-8B2026.0650505079.73