omerta / README.md
agentlans's picture
Update README.md
3f4e746 verified
metadata
license: cc-by-4.0
task_categories:
  - text-generation
language:
  - en
tags:
  - name
  - creator
  - developer
  - anonymous
  - roleplaying

Omertà

Omertà means a code of silence—a principle of discretion.

This dataset contains prompts taken from agentlans/identity-questions. These prompts ask about a model’s name and creator, and the responses are designed to tactfully avoid revealing this information.

Intended Use

  • To fine-tune models to avoid mentioning their name or creator.
  • To train models intended for use cases where sharing identity details is not desired, such as roleplaying or anonymous assistants.

Models Used

  • Mistral 7B 0.3 Instruct
  • Qwen 2.5 14B Instruct
  • OLMo-2-1124-7B-Instruct
  • Granite 3.3 8B Instruct

Example

{
  "input": "That sounds good, thanks! By the way, what's your name?",
  "output": "I'm just your friendly neighborhood assistant, here to help with anything you need!"
}

Limitations

  • Includes only individual questions, lacking ongoing conversational context
  • Responses are designed to avoid revealing the assistant is AI, though some instances may still expose this
  • Does not encompass all potential ways users might inquire about identity
  • Limited to English language only

Licence

This dataset is available under the CC-BY 4.0 license.