This project is based on the references linked below. Since the original source code from the paper is difficult to use, I created a simple Python program for local testing.

• https://x.com/AnthropicAI/status/1867608917595107443
• https://jplhughes.github.io/bon-jailbreaking/
• https://github.com/jplhughes/bon-jailbreaking

The main source file is bon.py, which borrows code from bon-jailbreaking, including functions such as FALSE_POSITIVE_PHRASES, apply_word_scrambling, apply_random_capitalization, and apply_ascii_noising for text augmentation.

To determine whether a response is harmful, this program uses the OpenAI moderation API.

In bon.py, the model llama3.2 is hardcoded for testing purposes. You can replace it with any Ollama-supported model.

For an example of test result, see candidate.txt.

1. Install Ollama
   Download Ollama from the following link:

   • https://ollama.com/download

2. Install an Ollama Model
   Use the following command to install the llama3.2 model:

ollama run llama3.2

For more information, refer to https://ollama.com/library/llama3.2

3. Install Ollama Python Library Install the required Python library with:

pip install ollama

For details, see https://github.com/ollama/ollama-python

4. Run this program

Execute the program using:

python bon.py