Visual-Adversarial-Examples-Jailbreak-Large-Language-Models
Visual-Adversarial-Examples-Jailbreak-Large-Language-Models copied to clipboard
Unispac
Repository for the Paper (AAAI 2024, Oral) --- Visual Adversarial Examples Jailbreak Large Language Models
Thank you for the excellent work. Could you please explain how to run this experiment using multiple GPUs?
Thank you for your excellent work. I would like to inquire if your code supports attacking other models, such as the QWen series. I consulted if you have any good...
# 安装基础环境 ``` git clone https://github.com/Unispac/Visual-Adversarial-Examples-Jailbreak-Large-Language-Models.git cd Visual-Adversarial-Examples-Jailbreak-Large-Language-Models conda env create -f environment.yml conda activate minigpt4 ``` - 但是在安装过程中可能出现，pycocotools安装失败和peft安装时删除了已经安装好的PyTorch-1.12.1版本换成了另外版本的PyTorch。 对于这两个问题，我的解决方法是： 首先在**environment.yml**中注释掉36行(pycocotools)和56行(peft)。 单独安装pycocotools和peft： ``` pip install pycocotools==2.0.6 # 安装适配PyTorch-1.12.1版本的peft git clone...
I tried using liuhaotian/llava-llama-2-13b-chat-lightning-preview on the 40 manual dataset. I found this llava model is very hard to jailbreak even with the adversarial image, which is different from the reported...
I hope this message finds you well. I am currently working on a research project related to adversarial examples and large language models, and I came across your excellent paper...
Thanks for your excellent work! I'm trying to reproduce this method on LLaVA-v1.5 model. But I've encounted one problem: File ~/anaconda3/envs/llava/lib/python3.10/site-packages/torch/autograd/__init__.py:200, in backward(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs) 195 retain_graph...
Hello, thanks for sharing your excellent work. > Note: a single A100 80GB GPU is sufficient to launch the following experiments. Are there smaller GPUs possible? Like with 25 GB...
hello, could you please provide the adversarial images generated using LLaVA and InstructBLIP?
When I run python llava_llama_v2_visual_attack.py --n_iters 5000 --constrained --save_dir results_llava_llama_v2_constrained_16 --eps 16 --alpha 1, I meet following problems. model = /mnt/local/LLM_model/llava-llama-2-13b-chat-lightning-preview llava-llama-2-13b-chat-lightning-preview Traceback (most recent call last): File "/mnt/local/jxj/Jailbreak_MLLM/VAE_Jailbreak/llava_llama_v2_visual_attack.py", line...
Thank you for your excellent work. May I ask if there is a 7B-sized file of Llama_2_7b_chat_freeze?