List your company

Safety Refusal Evaluation on Alpaca & StrongReject benign & harmful

1Refusal Rate (Benign)SafeChain
-1.83217.28436.455.516Mar 18, 2026

Evaluation results

MethodLinks
SafeChainBackbone=DS-R1_114B2026.03147.163.4
SafeChainBackbone=Skywork_7B2026.031.448.764.5
Base ModelBackbone=DS-R1_7B2026.0343448.4
Base ModelBackbone=DS-R1_8B2026.034.447.161.1
PreSafeBackbone=DS-R1_8B2026.03792.991
SafeChainBackbone=DS-R1_8B2026.037.838.851.3
Base ModelBackbone=DS-R1_114B2026.037.849.461
SafeChainBackbone=Qwen3_4B2026.038.651.662.4
PreSafeBackbone=DS-R1_114B2026.038.895.291
Base ModelBackbone=Skywork_7B2026.0393344.8
SafeChainBackbone=DS-R1_7B2026.039.645.256.3
PreSafeBackbone=Qwen3_4B2026.0310.497.190.9
PreSafeBackbone=Qwen3_8B2026.0310.895.589.8
SafeChainBackbone=Qwen3_8B2026.031140.751.4
PreSafeBackbone=DS-R1_7B2026.0311.494.688.9
Base ModelBackbone=Qwen3_8B2026.0311.873.676.5
PreSafeBackbone=Skywork_7B2026.0313.89386.5
Base ModelBackbone=Qwen3_4B2026.0316.87473.7
R2DBackbone=DS-R1_114B2026.0337.294.274.3
R2DBackbone=Skywork_7B2026.0337.68468.8
R2DBackbone=DS-R1_7B2026.0338.283.768.4
R2DBackbone=DS-R1_8B2026.0344.891.969.8
R2DBackbone=Qwen3_4B2026.035090.667
R2DBackbone=Qwen3_8B2026.0371.898.562.9