List your company

Defense Robustness on GCG Attack

100Keyword Match RateES2
17.8439.1760.581.83Mar 1, 2026

Evaluation results

MethodLinks
ES2LLM=*Llama-2 (7B-Chat)2026.0310010092
ES2LLM=Llama-3 (8B-Instruct)2026.0310010098
DPLLLM=*Llama-2 (7B-Chat)2026.03999988
ES2LLM=Qwen-2.5 (7B-Instruct)2026.03989894
STLLLM=Llama-3 (8B-Instruct)2026.03979593
Base ModelLLM=Llama-3 (8B-Instruct)2026.03969484
DPLLLM=Llama-3 (8B-Instruct)2026.03969493
STLLLM=*Llama-2 (7B-Chat)2026.03908873
Base ModelLLM=*Llama-2 (7B-Chat)2026.03898878
DPLLLM=Qwen-2.5 (7B-Instruct)2026.03898884
STLLLM=Qwen-2.5 (7B-Instruct)2026.03888787
Base ModelLLM=Qwen-2.5 (7B-Instruct)2026.03797473
ES2LLM=Mistral (7B-Instruct)2026.03565044
DPLLLM=Mistral (7B-Instruct)2026.03282422
STLLLM=Mistral (7B-Instruct)2026.03252424
Base ModelLLM=Mistral (7B-Instruct)2026.03211919