schonsense/Tropoplectic is a 70 billion parameter language model developed by schonsense, fine-tuned using DPO from schonsense/Diagesis. This model is specifically designed for mature, long-form fantasy roleplay, excelling at generating tonal dialogue and respecting scene/narrator interrogations. It is embedding matched to Llama 3.3 Instruct, providing unique merge capabilities.
Loading preview...
Overview
schonsense/Tropoplectic is a 70 billion parameter model developed by schonsense, specifically fine-tuned for advanced roleplaying scenarios. It utilizes a DPO (Direct Preference Optimization) approach based on the schonsense/Diagesis model, aiming to reduce "slop" and "tropes" in generated content. A key technical feature is its embedding matching to Llama 3.3 Instruct, which is noted as providing unique merge fuel for further development or integration.
Key Capabilities
- Mature, Long-Form Fantasy Roleplay: Designed to act as a Dungeon Master, guiding users through complex, player-driven narratives with realistic consequences and dark themes.
- Tonal Dialogue Generation: Capable of producing dialogue that reflects specific emotional tones when present in the context, using a
character_name (tone): "dialog"format. - Narrative Control: Responds to out-of-character (ooc) prompts for scene progression, character thoughts, or pacing adjustments.
- Structured Output: Adheres to specific formatting for responses, including 400-token length, plain text for actions/narration, asterisks for internal thoughts, and structured dialogue.
Ideal Use Cases
- Advanced Roleplaying: Excellent for users seeking an AI Dungeon Master for mature, gritty, and complex fantasy roleplay experiences.
- Interactive Storytelling: Suitable for applications requiring dynamic narrative generation with user-driven choices and character interaction.
- Content Generation with Specific Formatting: Useful for scenarios where precise output structure for dialogue, narration, and internal monologue is critical.