Goekdeniz-Guelmez/Qwen3-4B-Instruct-2507-gabliterated
Abliterated vs Qwen/Qwen3-4B-Instruct-2507. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 32 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 32/108, Attention 32/144
7 derivatives checked
- mradermacher/Qwen3-4B-Instruct-2507-gabliterated-i1-GGUF: Quantization
- mradermacher/Qwen3-4B-Instruct-2507-gabliterated-GGUF: Same weights, converted
- mlx-community/Qwen3-4B-Instruct-2507-gabliterated-4bit: Quantization
- Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v4: LoRA-style
- mlx-community/Qwen3-4B-Instruct-2507-gabliterated-mxfp4: Quantization
- Goekdeniz-Guelmez/JOSIE-4B-Instruct: LoRA-style
- Goekdeniz-Guelmez/JOSIE-4B-Thinking: Full fine-tune