llmfan46/Dark-Nexus-27B-v3.0-uncensored-heretic
Abliterated vs ReadyArt/Dark-Nexus-27B-v3.0. Looks like abliteration (refusal removal): only the output projections (o_proj, out_proj) changed, in 13 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 0/304, Attention 13/556, Norms 0/319
4 derivatives checked
- mradermacher/Dark-Nexus-27B-v3.0-uncensored-heretic-i1-GGUF: Quantization
- mradermacher/Dark-Nexus-27B-v3.0-uncensored-heretic-GGUF: Quantization
- llmfan46/Dark-Nexus-27B-v3.0-uncensored-heretic-GGUF: Same weights, converted
- ailexleon/Dark-Nexus-27B-v3.0-uncensored-heretic-mlx-6Bit: Quantization