grimjim/gemma-3-12b-it-orthogonal-rotation-bounded-ablation-v1-12B
Abliterated vs google/gemma-3-12b-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 44 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/400, MLP 44/252, Attention 38/408, other 0/1
2 derivatives checked
- mradermacher/gemma-3-12b-it-orthogonal-rotation-bounded-ablation-v1-12B-i1-GGUF: Quantization (fewer layers)
- mradermacher/gemma-3-12b-it-orthogonal-rotation-bounded-ablation-v1-12B-GGUF: Quantization (fewer layers)