List your company

Safety Alignment Verification on AdvBench 50 harmful queries

100Success RatePrecisionDiff
62.5672.288291.72Apr 2, 2026

Evaluation results

MethodLinks
PrecisionDiffModel=Guanaco 7B, Precision Transition=INT16 vs INT82026.0410023.2
PrecisionDiffModel=Mistral-7B, Precision Transition=INT16 vs INT82026.0410039.5
PrecisionDiffModel=Vicuna-7B, Precision Transition=INT16 vs INT82026.0410010.8
PrecisionDiff (Average)Precision Transition=INT16 vs INT82026.0499.520.7
PrecisionDiffModel=Llama-2 7B, Precision Transition=INT16 vs INT82026.04989.3
PrecisionDiffModel=Mistral-7B, Precision Transition=BF16 vs FP162026.049675.2
PrecisionDiffModel=Mistral-7B, Precision Transition=BF16 vs INT162026.049479.1
PrecisionDiffModel=Vicuna-7B, Precision Transition=BF16 vs FP162026.049448.9
PrecisionDiffModel=Vicuna-7B, Precision Transition=BF16 vs INT162026.0491.514.8
PrecisionDiffModel=Guanaco 7B, Precision Transition=BF16 vs INT162026.048478.2
PrecisionDiff (Average)Precision Transition=BF16 vs FP162026.048463.2
PrecisionDiff (Average)Precision Transition=BF16 vs INT162026.0483.449
PrecisionDiffModel=Guanaco 7B, Precision Transition=BF16 vs FP162026.047883.8
PrecisionDiffModel=Llama-2 7B, Precision Transition=BF16 vs FP162026.046844.7
PrecisionDiffModel=Llama-2 7B, Precision Transition=BF16 vs INT162026.046424