Nabbers1999/Gemma-3-27B-it-NP-Abliterated
Abliterated vs google/gemma-3-27b-it. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 21 of 62 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/4, Norms 0/484, MLP 21/294, Attention 21/464, other 0/1
2 derivatives checked
- mradermacher/Gemma-3-27B-Derestricted-i1-GGUF: Quantization (fewer layers)
- mradermacher/Gemma-3-27B-Derestricted-GGUF: Quantization (fewer layers)