coder3101/gemma-4-26B-A4B-it-qat-q4_0-unquantized-heretic
Abliterated vs google/gemma-4-26B-A4B-it-qat-q4_0-unquantized. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 9 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 0/231, Norms 0/433, other 0/122, Attention 9/223
4 derivatives checked
- SC117/gemma-4-26B-A4B-it-qat-heretic-GGUF: Same weights, converted
- mradermacher/gemma-4-26B-A4B-it-qat-q4_0-unquantized-heretic-i1-GGUF: Quantization
- mradermacher/gemma-4-26B-A4B-it-qat-q4_0-unquantized-heretic-GGUF: Quantization
- pegasus912/gemma-4-26b-a4b-it-qat-heretic-ud-q4-k-xl: Quantization