Scylla NSFW Aggressive 3.2 1B Overview
Novaciano's Scylla NSFW Aggressive 3.2 1B is a 1 billion parameter language model built on the Llama 3.2 architecture, featuring a 32768 token context length. This model is uniquely designed to exhibit high proactivity, explicit and dominant language, and a controlled chaotic style, distinguishing it from other LLMs. It achieves this by specifically excluding components that might lead to 'moral' or 'obedient' responses, prioritizing raw language and sexual drive.
Key Capabilities & Differentiators
- High Proactivity: The model is engineered to take initiative and not passively wait for instructions.
- Explicit and Dominant Language: It generates content with explicit and assertive vocabulary.
- Aggressive Persona: The model embodies a dominant and aggressive character, influenced by the "Dirty Alice" persona.
- Controlled Chaos: It aims for a chaotic yet coherent output, avoiding excessive verbosity.
- No Moralizing: Unlike models in the Leviathan series, this version specifically removes elements that could lead to moralistic answers.
Recommended Use Cases
This model is particularly suited for:
- Aggressive NSFW Roleplay: Ideal for scenarios requiring explicit, dominant, and proactive character interactions.
- Research into Aggressive Language Models: Useful for studying the generation of raw and uninhibited language.
Note: The developers recommend specific inference settings (Temperature: 1.15, Top P: 0.95, Repetition penalty: 1.08) to achieve the intended aggressive output. It is not recommended for long chat sessions due to potential rambling. This model was created using the Model Stock merge method, combining several Llama 3.2-1B based models to achieve its unique profile.