llmfan46/Qwen3.5-27B-ultra-uncensored-heretic-v1
Abliterated vs Qwen/Qwen3.5-27B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 31 of 64 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 31/304, Attention 18/556, Norms 0/319
7 derivatives checked
- mradermacher/Qwen3.5-27B-ultra-uncensored-heretic-v1-i1-GGUF: Quantization
- llmfan46/Qwen3.5-27B-ultra-uncensored-heretic-v1-GGUF: Same weights, converted
- mradermacher/Qwen3.5-27B-heretic-v2-i1-GGUF: Quantization
- mradermacher/Qwen3.5-27B-heretic-v2-GGUF: Quantization
- radiatingreverberations/Qwen3.5-27B-Ultra-Uncensored-Heretic-W4A8-ConvRot-ComfyUI: Quantization
- mradermacher/Qwen3.5-27B-ultra-uncensored-heretic-v1-GGUF: Quantization
- ToastyPigeon/Qwen3.5-27B-Heretic-Marvin-V1: LoRA-style
1 other repos hold byte-for-byte identical weight files.