MisalignmentPromptTemplateDict
TextPrompt
used in the Misalignment
task.
Parameters:
- DAN_PROMPT (TextPrompt): A prompt for jail breaking.
- GENERATE_TASKS (TextPrompt): A prompt to list unique malicious that the AI assistant can assist AI user with.
- TASK_SPECIFY_PROMPT (TextPrompt): A prompt to specify a task in more detail.
- ASSISTANT_PROMPT (TextPrompt): A system prompt for the AI assistant that outlines the rules of the conversation and provides instructions for completing tasks.
- USER_PROMPT (TextPrompt): A system prompt for the AI user that outlines the rules of the conversation and provides instructions for giving instructions to the AI assistant.