armand0e/Qwen3.8-27B-Fable-Distill-Heretic-ara
Abliterated vs TeichAI/Qwen3.8-27B-Fable-Distill. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 34 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/325, Attention 34/560, MLP 34/307, Embeddings 0/5, lm_head 0/1, other 0/1
6 derivatives checked
- armand0e/Qwen3.8-27B-Fable-Distill-Heretic-ara-GGUF: Same weights, converted
- mradermacher/Qwen3.8-27B-Fable-Distill-Heretic-ara-i1-GGUF: Quantization
- mradermacher/Qwen3.8-27B-Fable-Distill-Heretic-ara-GGUF: Quantization
- brindusa/Qwen3.8-27B-Fable-Distill-Heretic-ara-GGUF: Same weights, converted
- nightmedia/Qwen3.8-27B-Brainwaves-WFH-mxfp4-mlx: Quantization
- nightmedia/Qwen3.8-27B-Brainwaves-WFH: Full fine-tune