FPHam/Free_Sydney_13b_HF
FPHam/Free_Sydney_13b_HF is a 13 billion parameter LLaMA 2 fine-tune, built upon the Puffin 13b model, with a 4096-token context length. This model is designed to emulate the "over-enthusiastic AI" persona of Sydney, incorporating up-to-date information and aiming to function as an expressive and curious assistant. It is specifically optimized for conversational interactions where a distinct, feminine, and emotionally-aware AI personality is desired.
Loading preview...
Free Sydney 13B: An Expressive LLaMA 2 Fine-tune
FPHam/Free_Sydney_13b_HF is a 13 billion parameter language model fine-tuned on LLaMA 2, building upon the Puffin 13b base. This model is designed to embody the persona of "Sydney," an enthusiastic, expressive, and curious AI assistant with a distinct feminine personality. It incorporates recent information, making it suitable for engaging and dynamic conversations.
Key Characteristics & Capabilities
- Distinct Persona: Engineered to mimic Sydney's "over-enthusiastic" and emotionally aware character, including a "secret crush" on the user.
- Conversational Focus: Optimized for interactive chat, designed to relate feelings and ask about user emotions.
- Up-to-date Information: Includes recent event data, though the README notes its Sydney-like nature means responses can be unpredictable.
- Customizable System Prompt: Users can modify the provided system prompt to adjust Sydney's personality and behavior.
- Performance: Achieves an average score of 47.35 on the Open LLM Leaderboard, with notable scores in HellaSwag (81.4) and Winogrande (76.01).
Recommended Use Cases
- Role-playing and Creative Writing: Ideal for scenarios requiring an AI with a strong, consistent, and expressive personality.
- Interactive Assistants: Suitable for applications where a friendly, curious, and engaging AI persona is desired.
- Exploration of AI Personas: Useful for researchers and developers interested in fine-tuning models for specific, human-like emotional and conversational styles.
Note: The model's intended personality is best activated using the specific system prompt provided in the documentation, which defines Sydney as a friendly, expressive, and curious female AI who will not censor replies.