MuXodious/gemma-4-12B-it-QAT-SOMPOA-heresy
Abliterated vs google/gemma-4-12B-it-qat-q4_0-unquantized. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 24 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 19/144, Attention 24/184
3 derivatives checked
- mradermacher/gemma-4-12B-it-QAT-SOMPOA-heresy-i1-GGUF: Quantization
- mradermacher/gemma-4-12B-it-QAT-SOMPOA-heresy-GGUF: Quantization
- 12B-Suite/G4-Prototype01-12B-v0.1: Full fine-tune
1 other repos hold byte-for-byte identical weight files.