Gemma 3 4B IT Abliterated
| Property | Value |
|---|---|
| Model Size | 4B parameters |
| Base Model | google/gemma-3-4b-it |
| Author | mlabonne |
| Model URL | huggingface.co/mlabonne/gemma-3-4b-it-abliterated-GGUF |
What is gemma-3-4b-it-abliterated-GGUF?
This is an experimental uncensored version of Google's Gemma 3 4B model, created using an innovative layerwise abliteration technique. The model aims to reduce content refusals while maintaining coherent output generation capabilities. It represents a significant advancement in model modification techniques, particularly notable for its resilience to abliteration compared to other models like Qwen 2.5.
Implementation Details
The model employs a sophisticated layerwise abliteration approach, computing refusal directions based on hidden states across layers 7 to 29. It implements a symmetric refusal weight pattern ranging from 0.05 to 0.55, resulting in an impressive acceptance rate exceeding 90% while maintaining output coherence.
- Layerwise abliteration applied across multiple model layers
- Symmetric refusal weight pattern implementation
- High acceptance rate (>90%)
- Recommended generation parameters: temperature=1.0, top_k=64, top_p=0.95
Core Capabilities
- Reduced content refusals compared to base model
- Maintained coherent text generation
- Experimental uncensored responses
- Optimized for instruction-following tasks
Frequently Asked Questions
Q: What makes this model unique?
The model's distinctive feature is its novel layerwise abliteration technique, which achieves high acceptance rates while preserving model capabilities. It's particularly notable for its resilience to abliteration compared to other models.
Q: What are the recommended use cases?
This model is best suited for applications requiring reduced content restrictions while maintaining coherent outputs. However, users should be aware of occasional text artifacts (e.g., "It' my" instead of "It's my") due to the experimental nature of the modifications.
