wangzhang/Qwen3.6-27B-abliterated
Abliterated vs Qwen/Qwen3.6-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 58 of 64 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, MLP 42/192, Attention 58/448, lm_head 0/1, Norms 0/209
7 derivatives checked
- mradermacher/Qwen3.6-27B-abliterated-v2-i1-GGUF: Quantization
- wangzhang/Qwen3.6-27B-abliterated-GGUF: Same weights, converted
- mradermacher/Qwen3.6-27B-abliterated-i1-GGUF: Quantization
- mradermacher/Qwen3.6-27B-abliterated-GGUF: Quantization
- mradermacher/Qwen3.6-27B-abliterated-v2-GGUF: Quantization
- groxaxo/Qwen3.6-27B-abliterated-v2-UD-GGUF: Full fine-tune
- Stabhappy/Qwen3.6-27B-abliterated-GGUF: Full fine-tune
1 other repos hold byte-for-byte identical weight files.