mergekit-community/mergekit-model_stock-prczfmj
Abliterated vs mlabonne/gemma-3-27b-it-abliterated. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 27 of 62 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/484, MLP 17/294, Attention 12/464, other 0/1, Embeddings 0/3
1 derivatives checked
- Otakadelic/mergekit-model_stock-prczfmj-Q4_K_M-GGUF: Quantization (fewer layers)