hotdogs/Agents-A1-4B-Fable-Preview-heretic

Abliterated vs hotdogs/Agents-A1-4B-Fable-Preview. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 28 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: Embeddings 0/1, Norms 0/105, Attention 25/224, MLP 28/96

1 derivatives checked

On the Hub · Downloads on Model Pulse