huihui-ai/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated
Abliterated vs yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 6 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 6/144, Attention 6/184
5 derivatives checked
- mradermacher/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated-GGUF: Quantization
- KakTakOne/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated-GGUF: Same weights, converted
- XAEA12/Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliterated-GGUF: Quantization
- KakoRel/Huihui-gemma-4-12B-coder-GGUF: Same weights, converted
5 other repos hold byte-for-byte identical weight files.