huihui-ai/Huihui-granite-4.1-3b-abliterated
Abliterated vs ibm-granite/granite-4.1-3b. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 40 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 40/120, Attention 40/160
2 derivatives checked
- mradermacher/Huihui-granite-4.1-3b-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-granite-4.1-3b-abliterated-GGUF: Same weights, converted