SC117/Ornith-1.0-35B-Heretic-MTP
Abliterated vs ornith-ai/Ornith-1.0-35B. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 37 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/241, Attention 37/388, MLP 31/2299
4 derivatives checked
- SC117/Ornith-1.0-35B-Heretic-MTP-APEX-GGUF: Same weights, converted
- mradermacher/Ornith-1.0-35B-Heretic-MTP-i1-GGUF: Quantization
- mradermacher/Ornith-1.0-35B-Heretic-MTP-GGUF: Quantization
- ursb01/Ornith-1.0-35B-Heretic-MTP-APEX-GGUF: Same weights, converted