LiamVisionary/swarm-sovereign-12b
Abliterated vs google/gemma-4-12B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 20 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/12, Norms 0/289, other 0/48, MLP 20/144, Attention 6/184
2 derivatives checked
- mradermacher/swarm-sovereign-12b-i1-GGUF: Quantization
- mradermacher/swarm-sovereign-12b-GGUF: Quantization