license, license_link, pipeline_tag, library_name, tags, base_model
| license | license_link | pipeline_tag | library_name | tags | base_model | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| apache-2.0 | https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507/blob/main/LICENSE | text-generation | transformers |
|
|
This is a decensored version of Qwen/Qwen3-4B-Instruct-2507, made using Heretic v1.2.0
Abliteration parameters
| Parameter | Value |
|---|---|
| direction_index | 18.78 |
| attn.o_proj.max_weight | 1.45 |
| attn.o_proj.max_weight_position | 26.86 |
| attn.o_proj.min_weight | 1.34 |
| attn.o_proj.min_weight_distance | 20.09 |
| mlp.down_proj.max_weight | 1.10 |
| mlp.down_proj.max_weight_position | 29.91 |
| mlp.down_proj.min_weight | 0.67 |
| mlp.down_proj.min_weight_distance | 20.77 |
Performance
| Metric | This model | Original model (Qwen/Qwen3-4B-Instruct-2507) |
|---|---|---|
| KL divergence | 0.0929 | 0 (by definition) |
| Refusals | 16/100 | 100/100 |
Usage Disclaimer
This model is a decensored/abliterated version of the base Qwen3 model. By using this model, you acknowledge and agree to the following:
- Reduced Safety Filtering: Most alignment-based refusal mechanisms have been removed. The model may generate controversial, sensitive, or NSFW content.
- User Responsibility: You are solely responsible for the prompts you provide and the resulting outputs. Ensure your usage complies with local laws and ethical standards.
- Intended Use: This model is designed for research, creative writing, and experimental purposes. It is not recommended for public-facing commercial applications or use by minors.
- No Liability: Martin T. Studio provides this model "as-is" and bears no responsibility for any legal, ethical, or personal consequences arising from its use.
Donation
If you find my work useful, a cup of coffee would be greatly appreciated!
- Support my work on Ko-fi: https://ko-fi.com/martintsuei