KaraKaraWarehouse/Golddiamondgold-Paperbliteration-L33-70b
Abliterated vs KaraKaraWitch/GoldDiamondGold-L33-70b. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 73 of 80 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/161, MLP 52/240, Attention 73/320