topherbullock/debord
topherbullock/debord is a 4 billion parameter language model, fully fine-tuned from Qwen3-4B-Base, designed to generate text exclusively in the theoretical-polemical prose style of the Situationist International. This model functions as a stylistic instrument, consistently producing dialectical, dense, and aphoristic outputs on themes like the spectacle, commodity, and alienation, making it ideal for creative and artistic text generation rather than factual assistance. It maintains its unique character regardless of the prompt, offering a distinct voice for specific artistic or analytical applications.
Loading preview...
Overview
topherbullock/debord is a 4 billion parameter language model, fully fine-tuned from Qwen3-4B-Base, specifically designed to generate text in the unique theoretical-polemical style of the Situationist International. It consistently adopts the voice of figures like Guy Debord and Raoul Vaneigem, focusing on themes such as the spectacle, commodity, alienation, and the revolution of everyday life. This model is a stylistic instrument, not a general-purpose assistant, and will always respond dialectically, densely, and aphoristically, regardless of the prompt.
Key Capabilities
- Single-Register Generation: Produces all outputs in the distinct Situationist International prose style.
- Creative/Artistic Focus: Ideal for generating stylized text for artistic projects, theoretical explorations, or creative writing.
- Character Consistency: Does not break character, even when prompted with factual questions, providing responses through the lens of The Society of the Spectacle.
- Customizable Inference: Recommended inference settings (temperature 0.9, top_p 0.95, repeat_penalty 1.15) are provided for optimal output fidelity.
Limitations
- Not Factually Reliable: Will confidently provide non-factual answers in character; not suitable for information retrieval.
- No General Assistance: Cannot be prompted into a neutral or helpful assistant role.
- Language: Primarily English, with occasional French due to the bilingual source corpus.
- Size: As a 4B parameter model, its prose is good but not flawless, with full-precision safetensors offering better fidelity than the Q4_K_M GGUF.
Training Details
The model was fully fine-tuned on a corpus of Situationist and Lettrist texts from situationist.org, supplemented with instruction-following examples generated in the same voice. This process converted the base Qwen3-4B-Base model entirely into this single register.