ewald1976/findesiecle-12b
The ewald1976/findesiecle-12b is a 12 billion parameter creative writing fine-tune of Mistral Nemo 12B, developed by ewald1976. It is specifically trained on a curated literary dataset in the English contemplative tradition, emphasizing stylistic restraint and psychological subtext. This model excels at generating narrative and roleplay content with a period sensibility, focusing on observation and economy of language rather than dramatic intensity. It is best suited for users seeking thoughtful, nuanced prose in creative writing and roleplay scenarios.
Loading preview...
Overview
findesiecle-12b is a 12 billion parameter model, fine-tuned from Mistral Nemo 12B, specifically designed for creative writing and narrative generation. Its training dataset comprises curated literary works from the English contemplative tradition, including authors like Huxley, Wells, and Shelley, fostering a unique stylistic output. The model prioritizes stylistic restraint, economy of language, and psychological subtext over dramatic intensity or explicit content.
Key Capabilities & Characteristics
- Creative Writing & Roleplay: Generates nuanced narratives and roleplay outputs.
- Contemplative Style: Emphasizes observation, quiet introspection, and period sensibility.
- Stylistic Restraint: Avoids verbose descriptions and dramatic intensity, focusing on conveying meaning through subtle prose.
- Psychological Subtext: Characters reveal themselves through actions and unspoken thoughts rather than explicit declarations.
- Period Sensibility: Comfortable generating content aligned with 19th and early 20th-century English literary styles.
Intended Use Cases
This model is ideal for users who appreciate prose that:
- Values economy and observation over explanation.
- Describes uncanny or remarkable events with sachlichkeit (calm objectivity).
- Prefers psychological depth and interiority in characters.
- Seeks a model that can write scenes where "nothing dramatic happens and yet something is conveyed."
Not Recommended For
findesiecle-12b is explicitly not designed for generating explicit NSFW content, shock value, graphic violence, or aggressive uncensored output. Its training dataset contains no NSFW material, and it naturally leans away from such themes.
Recommended Inference Settings
For optimal results, the model performs best with lower temperatures to maintain its characteristic restraint:
- Temperature: 0.6 – 0.8
- Top-p: 0.9
- Repetition penalty: 1.05 – 1.10