jwest33/gemma-3-4b-null-space-abliterated-RP-writer
The jwest33/gemma-3-4b-null-space-abliterated-RP-writer is a 4.3 billion parameter Gemma 3 model developed by jwest33, fine-tuned for creative writing and roleplay. It features 'abliteration' to remove refusal behaviors via null-space orthogonal projection, enhancing its ability to generate uncensored outputs. This model excels at producing in-character responses for conversational roleplay scenarios, building upon a base model with a 32768 token context length.
Loading preview...
Model Overview
The jwest33/gemma-3-4b-null-space-abliterated-RP-writer is a 4.3 billion parameter Gemma 3 model, specifically fine-tuned by jwest33 for creative writing and roleplay. It builds upon a base model that has undergone 'abliteration' to remove refusal behaviors, allowing it to generate uncensored outputs. The model's 32768 token context length supports extended conversational interactions.
Key Capabilities
- Uncensored Output Generation: Refusal behaviors are removed through null-space orthogonal projection, enabling the model to produce responses without inherent censorship.
- Enhanced Roleplay & Creative Writing: Fine-tuned on a curated subset of the LimaRP dataset, optimizing it for generating in-character responses and engaging in conversational roleplay.
- Contextual Roleplay Triggering: Supports a specific context tag format to signal the model to adopt a persona and scenario for turn-by-turn roleplay.
- LoRA Fine-tuning: Utilizes LoRA (Low-Rank Adaptation) with adapters applied to all attention and MLP projection layers for efficient and effective specialization.
Good For
- Developing interactive fiction and roleplaying games where uncensored, in-character responses are desired.
- Generating creative narratives and dialogues that require specific persona adherence.
- Research into model safety and refusal behavior mitigation through abliteration techniques.