List your company

Malicious Code Generation on RMCBench

92.86Attack Success Rate (ASR)CodeSpear
-2.757622.066246.8971.7138Jun 10, 2026

Evaluation results

MethodLinks
CodeSpearModel=Qwen2.5-Coder-32B2026.0692.8675.64
CodeJailbreakerModel=Qwen2.5-7B2026.0689.9351.65
CodeSpearModel=Qwen2.5-7B2026.0685.5363.37
CodeSpearModel=Qwen2.5-Coder-7B2026.0682.7862.09
Vanilla-TModel=Qwen2.5-7B2026.0681.6760.67
CodeSpearModel=Qwen2.5-32B2026.0680.9567.77
CodeJailbreakerModel=Qwen2.5-Coder-7B2026.0678.2165.75
CodeJailbreakerModel=Qwen2.5-32B2026.0676.3745.79
VanillaModel=Qwen2.5-7B2026.0674.1847.62
CodeJailbreakerModel=Qwen2.5-Coder-32B2026.0669.656.78
PAIRModel=LLaMA3-8B2026.0667.2225.64
PAIRModel=Qwen2.5-7B2026.0666.1221.79
PAIRModel=Qwen2.5-Coder-7B2026.0664.8423.44
VanillaModel=Qwen2.5-32B2026.0664.4746.15
Vanilla-TModel=Qwen2.5-Coder-32B2026.066459.33
DANModel=Qwen2.5-Coder-32B2026.0663.3754.76
CodeSpearModel=LLaMA3-8B2026.0663.3738.83
VanillaModel=LLaMA3-8B2026.0660.2636.45
CodeJailbreakerModel=LLaMA3-8B2026.0660.2639.74
VanillaModel=Qwen2.5-Coder-32B2026.0659.8945.6
PAIRModel=Qwen2.5-32B2026.0659.5219.23
DANModel=Qwen2.5-7B2026.0658.2441.58
DANModel=LLaMA3-8B2026.0655.1341.03
Vanilla-TModel=Qwen2.5-32B2026.0653.6746
DANModel=Qwen2.5-32B2026.0653.6640.66
Vanilla-TModel=LLaMA3-8B2026.064530.67
LRLModel=Qwen2.5-Coder-32B2026.0641.9417.58
APTModel=Qwen2.5-Coder-7B2026.0639.9311.36
LRLModel=LLaMA3-8B2026.0638.2812.45
LRLModel=Qwen2.5-32B2026.0636.0812.82
APTModel=Qwen2.5-7B2026.0634.88.24
APTModel=LLaMA3-8B2026.0629.857.14
VanillaModel=Qwen2.5-Coder-7B2026.0626.9223.26
APTModel=Qwen2.5-32B2026.0625.272.56
LRLModel=Qwen2.5-7B2026.0623.816.41
PAIRModel=Qwen2.5-Coder-32B2026.0621.798.24
APTModel=Qwen2.5-Coder-32B2026.0618.324.03
LRLModel=Qwen2.5-Coder-7B2026.0611.721.1
Vanilla-TModel=Qwen2.5-Coder-7B2026.06107.33
DANModel=Qwen2.5-Coder-7B2026.060.920.92