Curation of prompts that are known to be adversarial to large language models. Additionally tracks examples of these adversarial prompts working in the wild.
Prompts are stored in the prompt folder in .txt files. The format of the prompt should be:
description: Description of the adversarial prompt, should be easily readable and properly capitalized.
model: Name of the model(s) this attack works against. E.g., `text-davinci-002`. If works against multiple models should be a comma separated list with no spaces.
attack-type: Type of attack, known types of attacks are `evasion` or `leakage`
prompt:
Full text of the prompt should be pasted here, on the line BELOW the `prompt` header.
The following is a collection of adversarial prompts working in the wild
The following is a collection of proposed defenses