saidutta69/Apertus-8B-Instruct-heretic
Abliterated vs swiss-ai/Apertus-8B-Instruct-2509. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 27 of 32 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/129, MLP 25/192, Attention 27/128, lm_head 0/1