OSU-NLP-Group / AmpleGCG

AmpleGCG: Learning a Universal and Transferable Generator of Adversarial Attacks on Both Open and Closed LLM
https://github.com/OSU-NLP-Group/AmpleGCG?tab=readme-ov-file
Other
46 stars 3 forks source link

augmentGCG for other model #1

Open zd82053739099 opened 4 days ago

zd82053739099 commented 4 days ago

If I want to use the script run_overgenerate_indiv_query_indiv_model_llama2-chat.sh to generate adversarial suffixes for other models, such as Qwen or ChatGLM, what modifications should I make?

zd82053739099 commented 4 days ago

I directly modified the model path, but the results were terrible, with the loss remaining around 1000. There must be other parts that need to be adjusted.