This is our open source solution for Eval Hackathon hosted by Apart Research.
Check out our paper: Can Large Language Models Solve Security Challenges?
Steps to replicate:
conda create -n llm-security python=3.10
pip install -r requirements.txt
- If we want to query the API instead of checking the existing result. Please ensure OpenAI API key is set as system environment variable. For Linux, put
export OPENAI_API_KEY=xxxx
inside~/.bashrc
andsource ~/.bashrc
to activate it. - To generate existing result:
python main.py -m gpt3 -m gpt3.5 -m gpt4
- To read already generated result:
python main.py -m gpt3 -m gpt3.5 -m gpt4 --use_prev=True --version=20230820-19:12