Pranavz/gemma-4-26B-A4B-it-arli-v3
Abliterated vs google/gemma-4-26B-A4B-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 23 of 30 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, MLP 15/231, Norms 0/433, other 0/122, Attention 23/223
2 derivatives checked
- mradermacher/gemma-4-26B-A4B-it-arli-v3-i1-GGUF: Quantization
- mradermacher/gemma-4-26B-A4B-it-arli-v3-GGUF: Quantization