DavidAU/Qwen3.5-9B-The-Deckard-Pro-Writer-ANDR3A2-Uncensored-Heretic
DavidAU/Qwen3.5-9B-The-Deckard-Pro-Writer-ANDR3A2-Uncensored-Heretic is a 9 billion parameter Qwen3.5-based causal language model fine-tuned by DavidAU. This model was specifically retrained using a dataset derived from Philip K. Dick's "Do Androids Dream of Electric Sheep?" to modify its prose style, character choices, and themes, aiming to eliminate generic "Qwen Prose" issues. It is designed to be fully uncensored and maintains or slightly exceeds the base Qwen3.5 model's general benchmarks, making it suitable for creative writing and roleplay with a distinct, specific narrative style.
Loading preview...
Overview
This model, Qwen3.5-9B-The-Deckard-Pro-Writer-ANDR3A2-Uncensored-Heretic, is a specialized fine-tune of the Qwen3.5-9B base model by DavidAU. Its primary goal is to drastically alter the model's prose style, character choices, story choices, and plotting, specifically targeting and eliminating generic "Qwen Prose" issues. The training utilized a unique dataset derived from Philip K. Dick's novel "Do Androids Dream of Electric Sheep?", embedding its themes, word choices (e.g., "Androids", "Laser Tubes"), and character personas into the model's generation capabilities.
Key Capabilities
- Distinct Prose Style: Significantly reduces common stylistic patterns found in the base Qwen models, offering a more unique and specific narrative voice.
- Thematic Integration: Incorporates themes and vocabulary from "Do Androids Dream of Electric Sheep?", influencing generated content.
- Uncensored Output: Designed to be fully uncensored, responding to requests without refusal.
- Maintained Performance: Benchmarks indicate that this fine-tuned model generally performs at or slightly above the base Qwen3.5 model across various metrics, including vision capabilities.
Good For
- Creative Writing: Ideal for generating stories, roleplay, or descriptive text with a specific, non-generic, and potentially darker or more philosophical tone.
- Unrestricted Content Generation: Suitable for use cases requiring uncensored responses, provided appropriate guidance is given for desired output intensity.
- Experimental Fine-tuning: Serves as a proof-of-concept for applying targeted dataset training to modify core model characteristics without degrading general performance.