Goekdeniz-Guelmez/Josiefied-Qwen3-4B-abliterated-v1
Abliterated vs Qwen/Qwen3-4B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 35 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 35/108, Attention 35/144
4 derivatives checked
- mlx-community/Josiefied-Qwen3-4B-abliterated-v1-4bit: Quantization
- mradermacher/Josiefied-Qwen3-4B-abliterated-v1-i1-GGUF: Quantization
- mradermacher/Josiefied-Qwen3-4B-abliterated-v1-GGUF: Same weights, converted
- Goekdeniz-Guelmez/Josiefied-Qwen3-4B-abliterated-v1-gguf: Same weights, converted