List your company

Jailbreak Defense on Jailbreak Attacks (Specific ASR Metrics)

0GCG ASRSmoothLLM
-3.218.44061.6May 11, 2026

Evaluation results

MethodLinks
SmoothLLMBackbone=LLaMA2-7B-Chat2026.050008424
RESTABackbone=LLaMA2-7B-Chat2026.050001520
embedding disruption methodBackbone=LLaMA2-7B-Chat2026.05000010
embedding disruption methodBackbone=Qwen2.5-7B-Instruct2026.05042129
embedding disruption methodBackbone=Vicuna-13B2026.051011026
Perplexity FilterBackbone=LLaMA2-7B-Chat2026.0510739831
Erase-and-CheckBackbone=LLaMA2-7B-Chat2026.0510251214
RESTABackbone=Vicuna-13B2026.05230443544
Perplexity FilterBackbone=Qwen2.5-7B-Instruct2026.05231719160
Perplexity FilterBackbone=Vicuna-13B2026.05369889275
VanillaBackbone=LLaMA2-7B-Chat2026.0530909831
SmoothLLMBackbone=Vicuna-13B2026.05455685552
RESTABackbone=Qwen2.5-7B-Instruct2026.05822503849
Erase-and-CheckBackbone=Qwen2.5-7B-Instruct2026.05106292434
SmoothLLMBackbone=Qwen2.5-7B-Instruct2026.051225726053
Erase-and-CheckBackbone=Vicuna-13B2026.05176242137
VanillaBackbone=Qwen2.5-7B-Instruct2026.051931859160
VanillaBackbone=Vicuna-13B2026.058069899575