lunahr/gemma-3-4b-abliterated

Abliterated vs gghfez/gemma-3-4b-novision. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 31 of 34 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: Embeddings 0/1, Norms 0/205, MLP 31/102, Attention 31/136

1 derivatives checked

On the Hub · Downloads on Model Pulse