MethodWhite/Qwen3.5-9B-Abliterated-HSAQ-v2
Abliterated vs Qwen/Qwen3.5-9B-Base. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 5 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Attention 5/224, Norms 0/105
2 derivatives checked
- mradermacher/Qwen3.5-9B-Abliterated-HSAQ-v2-i1-GGUF: Quantization
- mradermacher/Qwen3.5-9B-Abliterated-HSAQ-v2-GGUF: Same weights, converted