asfgsdfg/Qwen3.8-27B-Heretic
Abliterated vs Qwen/Qwen3.8-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 29 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/319, Attention 20/556, MLP 29/304, Embeddings 0/4, lm_head 0/1
2 derivatives checked
- mradermacher/Qwen3.8-27B-Heretic-i1-GGUF: Quantization
- mradermacher/Qwen3.8-27B-Heretic-GGUF: Quantization