huihui-ai/Huihui-gemma-4-12B-it-qat-q4_0-unquantized-abliterated
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 14 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 14/144, Attention 14/184
2 derivatives checked
- mradermacher/Huihui-gemma-4-12B-it-qat-q4_0-unquantized-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-gemma-4-12B-it-qat-q4_0-unquantized-abliterated-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.