Overview

BeaverAI/MN-2407-DSK-QwQify-v0.1-12B is a 12 billion parameter language model developed by BeaverAI, built upon a lineage starting with mistralai/Mistral-Nemo-Base-2407, then PocketDoc/Dans-PersonalityEngine-V1.1.0-12b, and finally PocketDoc/Dans-SakuraKaze-V1.0.0-12b. The primary goal of this fine-tune is to integrate "QwQ's thoughts" into the base model, enhancing its performance in specific conversational and narrative generation tasks. It maintains a substantial context length of 32768 tokens.

Key Capabilities

• Enhanced Roleplay & Co-writing: Specifically fine-tuned to improve an existing roleplay/adventure/co-writing model with a distinct "QwQ" persona.
• ChatML Formatting: Designed to work with ChatML, requiring users to remove thinking from previous turns and potentially add <think>\n for automatic thought generation.
• Customizable Character Prefix: Users can disable the default {{char}}: prefix and integrate custom instructions into the system prompt for more controlled character output.

Good for

• Interactive Storytelling: Ideal for applications requiring dynamic and engaging adventure or co-writing experiences.
• Character-Driven Roleplay: Suited for scenarios where the model needs to embody a specific character persona with consistent "thoughts" and dialogue.
• Developers Experimenting with Persona Injection: Useful for those looking to understand and implement specific personality traits into existing base models through fine-tuning.