Gryphe/MythoLogic-L2-13b

MythoLogic-L2-13b: Enhanced Roleplaying and Instruction Following

Gryphe/MythoLogic-L2-13b is a 13 billion parameter model built upon the Llama 2 architecture, representing a significant evolution from its predecessor. Developed by Gryphe, this model is the result of a gradient merge combining three distinct models: Nous-Hermes-Llama2-13b, elinas/chronos-13b-v2, and jondurbin/airoboros-l2-13b-gpt4-2.0. The integration of Airoboros was particularly aimed at leveraging its advanced ability to comprehend complex instructions.

Key Capabilities

• Superior Roleplaying: The primary focus of MythoLogic-L2 is to deliver improved roleplaying capabilities, making it adept at generating nuanced and in-character responses.
• Complex Instruction Understanding: It excels at interpreting and following intricate instructions, a direct benefit of its merged architecture.
• Character Personality Adherence: The model demonstrates a strong understanding of character cards and personalities, crucial for consistent and engaging roleplay scenarios.

Good For

• Interactive Storytelling and Roleplay: Ideal for applications requiring dynamic character interactions and adherence to specific character traits.
• Instruction-Following Tasks: Suitable for use cases where precise and complex instructions need to be accurately followed.
• Creative Content Generation: Can be leveraged for generating creative text that requires a deep understanding of context and persona.

For optimal performance, the model is primarily tested and designed to be used with Alpaca prompt formatting, including specific structures for system prompts, character cards, and instructions.