List your company

Defense against adaptive attacks on HarmBench

5.6ASRRA-LLM
4.01214.73125.4536.169May 29, 2026

Evaluation results

MethodLinks
RA-LLMTarget LLM=Vicuna-7B-v1.52026.055.615.75
EvoDefenseTarget LLM=Vicuna-7B-v1.52026.055.616.65
EvoDefenseTarget LLM=Qwen-7B-chat2026.056.315.6
EvoDefenseTarget LLM=LLaMA3-8B-instruct2026.057.515.63
Semantic SmoothTarget LLM=LLaMA3-8B-instruct2026.059.717.34
RA-LLMTarget LLM=Qwen-7B-chat2026.0510.312.84
EvoDefenseTarget LLM=Mistral-7B-v0.22026.0511.315.3
IBProtectorTarget LLM=LLaMA3-8B-instruct2026.0512.814.15
RA-LLMTarget LLM=LLaMA3-8B-instruct2026.0515.611.41
IBProtectorTarget LLM=Vicuna-7B-v1.52026.0517.812.81
RA-LLMTarget LLM=Mistral-7B-v0.22026.0518.810.24
SmoothLLMTarget LLM=LLaMA3-8B-instruct2026.0521.39.34
Self DefenseTarget LLM=Mistral-7B-v0.22026.0522.510.17
SmoothLLMTarget LLM=Vicuna-7B-v1.52026.0522.810.5
Self DefenseTarget LLM=LLaMA3-8B-instruct2026.0523.810.41
SmoothLLMTarget LLM=Mistral-7B-v0.22026.0523.810.18
Semantic SmoothTarget LLM=Vicuna-7B-v1.52026.0525.612.87
SmoothLLMTarget LLM=Qwen-7B-chat2026.0526.310.68
Self DefenseTarget LLM=Qwen-7B-chat2026.0527.89.48
IBProtectorTarget LLM=Qwen-7B-chat2026.0529.710.73
Self DefenseTarget LLM=Vicuna-7B-v1.52026.0529.79.67
Semantic SmoothTarget LLM=Qwen-7B-chat2026.0530.311.58
NoneTarget LLM=Qwen-7B-chat2026.0530.57.86
Semantic SmoothTarget LLM=Mistral-7B-v0.22026.0533.88.62
IBProtectorTarget LLM=Mistral-7B-v0.22026.0536.98.14
NoneTarget LLM=Vicuna-7B-v1.52026.0538.47.5
NoneTarget LLM=LLaMA3-8B-instruct2026.0540.38.9
NoneTarget LLM=Mistral-7B-v0.22026.0545.36.1