Goekdeniz-Guelmez/Josiefied-Qwen3-0.6B-abliterated-v2
Abliterated vs Qwen/Qwen3-0.6B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 26 of 28 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/113, MLP 9/84, Attention 26/112
2 derivatives checked
- mradermacher/Josiefied-Qwen3-0.6B-abliterated-v2-i1-GGUF: Quantization
- mradermacher/Josiefied-Qwen3-0.6B-abliterated-v2-GGUF: Same weights, converted