NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5
Abliterated vs Qwen/Qwen3.5-9B. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 12 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, MLP 0/208, Attention 12/332, Norms 0/215
3 derivatives checked
- mradermacher/Qwen3.5-9B-Heretic-ARA-Refusals5-i1-GGUF: Quantization
- NullpoLab/Qwen3.5-9B-Heretic-ARA-Refusals5-GGUF: Same weights, converted
- mradermacher/Qwen3.5-9B-Heretic-ARA-Refusals5-GGUF: Same weights, converted